File: index.html

package info (click to toggle)
nvidia-cuda-toolkit 11.8.0-5~deb12u1
links: PTS, VCS
area: non-free
in suites: bookworm
size: 18,338,396 kB
sloc: ansic: 172,472; cpp: 57,058; javascript: 21,597; python: 12,656; xml: 12,438; makefile: 2,949; sh: 2,056; perl: 352
file content (12601 lines) | stat: -rw-r--r-- 978,152 bytes
parent folder | download | duplicates (3)
<!DOCTYPE html
  PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" lang="en-us" xml:lang="en-us">
   <head>
      <meta http-equiv="Content-Type" content="text/html; charset=utf-8"></meta>
      <meta http-equiv="X-UA-Compatible" content="IE=edge"></meta>
      <meta name="copyright" content="(C) Copyright 2005"></meta>
      <meta name="DC.rights.owner" content="(C) Copyright 2005"></meta>
      <meta name="DC.Type" content="concept"></meta>
      <meta name="DC.Title" content="User Guide"></meta>
      <meta name="DC.Coverage" content="Nsight Systems"></meta>
      <meta name="DC.subject" content="User Guide, CLI, mpirun, qdstrm, GUI, macOS, Windows, Linux, Profiling, Ftrace, Context_Switch, Symbols, Hotkey, QNX, Arrow, Exporter, Export, Sqlite, JSON, Txt, Enum, Samples, Examples, Reports, Scripts, Formats, CSV, HTML, TSV, Statistics, nvprof, NVVP, Migration, Docker, Container, DX11, DX12, D3D11, D3D12, DirectX, WDDM, Vulkan, Stutter, OpenGL, Frame, VSYNC, FPS, OpenMP, OMPT, OSRT, Pthread, NVTX, Domain, Range, Category, CUDA, allocation, memory, graph, OpenACC, OpenGL_ES, ETW, Metric, Counter, PCIe, NVLink, network, communication, MPI, UCX, NCCL, OpenSHMEM, NVSHMEM, OpenMPI, MPICH, Report, nsys-rep, Qdproj, Timeline, Events, Function, Top_Down, Bottom_Up, Flat, Filtering, analyze, expert, stats, NVTXT, ImportNvtxt, Integration, Visual Studio, Troubleshooting, Debugging, Feedback, Launch, Backtrace, ELF, Blogposts, Videos, Training, Presentations, Support"></meta>
      <meta name="keywords" content="User Guide, CLI, mpirun, qdstrm, GUI, macOS, Windows, Linux, Profiling, Ftrace, Context_Switch, Symbols, Hotkey, QNX, Arrow, Exporter, Export, Sqlite, JSON, Txt, Enum, Samples, Examples, Reports, Scripts, Formats, CSV, HTML, TSV, Statistics, nvprof, NVVP, Migration, Docker, Container, DX11, DX12, D3D11, D3D12, DirectX, WDDM, Vulkan, Stutter, OpenGL, Frame, VSYNC, FPS, OpenMP, OMPT, OSRT, Pthread, NVTX, Domain, Range, Category, CUDA, allocation, memory, graph, OpenACC, OpenGL_ES, ETW, Metric, Counter, PCIe, NVLink, network, communication, MPI, UCX, NCCL, OpenSHMEM, NVSHMEM, OpenMPI, MPICH, Report, nsys-rep, Qdproj, Timeline, Events, Function, Top_Down, Bottom_Up, Flat, Filtering, analyze, expert, stats, NVTXT, ImportNvtxt, Integration, Visual Studio, Troubleshooting, Debugging, Feedback, Launch, Backtrace, ELF, Blogposts, Videos, Training, Presentations, Support"></meta>
      <meta name="DC.Format" content="XHTML"></meta>
      <meta name="DC.Identifier" content="abstract"></meta>
      <link rel="stylesheet" type="text/css" href="../common/formatting/commonltr.css"></link>
      <link rel="stylesheet" type="text/css" href="../common/formatting/site.css"></link>
      <title>User Guide :: Nsight Systems Documentation</title>
      <!--[if lt IE 9]>
      <script src="../common/formatting/html5shiv-printshiv.min.js"></script>
      <![endif]-->
      <script type="text/javascript" charset="utf-8" src="../common/scripts/tynt/tynt.js"></script>
      --&gt;
      
      <script src="https://assets.adobedtm.com/5d4962a43b79/c1061d2c5e7b/launch-191c2462b890.min.js"></script>
      <script type="text/javascript" charset="utf-8" src="../common/formatting/jquery.min.js"></script>
      <script type="text/javascript" charset="utf-8" src="../common/formatting/jquery.ba-hashchange.min.js"></script>
      <script type="text/javascript" charset="utf-8" src="../common/formatting/jquery.scrollintoview.min.js"></script>
      <script type="text/javascript" src="../search/htmlFileList.js"></script>
      <script type="text/javascript" src="../search/htmlFileInfoList.js"></script>
      <script type="text/javascript" src="../search/nwSearchFnt.min.js"></script>
      <script type="text/javascript" src="../search/stemmers/en_stemmer.min.js"></script>
      <script type="text/javascript" src="../search/index-1.js"></script>
      <script type="text/javascript" src="../search/index-2.js"></script>
      <script type="text/javascript" src="../search/index-3.js"></script>
      <link rel="canonical" href="https://docs.nvidia.com/nsight-systems/UserGuide/index.html"></link>
      <link rel="stylesheet" type="text/css" href="../common/formatting/qwcode.highlight.css"></link>
   </head>
   <body>
      
      <header id="header"><span id="company">NVIDIA</span><span id="site-title">Nsight Systems Documentation</span><form id="search" method="get" action="search">
            <input type="text" name="search-text"></input><fieldset id="search-location">
               <legend>Search In:</legend>
               <label><input type="radio" name="search-type" value="site"></input>Entire Site</label>
               <label><input type="radio" name="search-type" value="document"></input>Just This Document</label></fieldset>
            <button type="reset">clear search</button>
            <button id="submit" type="submit">search</button></form>
      </header>
      <div id="site-content">
         <nav id="site-nav">
            <div class="category closed"><a href="../index.html" title="The root of the site.">Nsight Systems
                  v2022.4.1</a></div>
            <div class="category"><a href="index.html" title="User Guide">User Guide</a></div>
            <ul>
               <li>
                  <div class="section-link"><a href="#cli-profiling">1.&nbsp;Profiling from the CLI</a></div>
                  <ul>
                     <li>
                        <div class="section-link"><a href="#cli-installing">1.1.&nbsp;Installing the CLI on Your Target</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#cli-options">1.2.&nbsp;Command Line Options</a></div>
                        <ul>
                           <li>
                              <div class="section-link"><a href="#cli-global-options">1.2.1.&nbsp;CLI Global Options</a></div>
                           </li>
                        </ul>
                     </li>
                     <li>
                        <div class="section-link"><a href="#cli-command-switches">1.3.&nbsp;CLI Command Switches</a></div>
                        <ul>
                           <li>
                              <div class="section-link"><a href="#cli-profile-command-switch-options">1.3.1.&nbsp;CLI Profile Command Switch Options</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#cli-start-command-switch-options">1.3.2.&nbsp;CLI Start Command Switch Options</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#cli-stop-command-switch-options">1.3.3.&nbsp;CLI Stop Command Switch Options</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#cli-cancel-command-switch-options">1.3.4.&nbsp;CLI Cancel Command Switch Options</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#cli-launch-command-switch-options">1.3.5.&nbsp;CLI Launch Command Switch Options</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#cli-shutdown-command-switch-options">1.3.6.&nbsp;CLI Shutdown Command Switch Options</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#cli-export-command-switch-options">1.3.7.&nbsp;CLI Export Command Switch Options</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#cli-stats-command-switch-options">1.3.8.&nbsp;CLI Stats Command Switch Options</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#cli-analyze-command-switch-options">1.3.9.&nbsp;CLI Analyze Command Switch Options</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#cli-status-command-switch-options">1.3.10.&nbsp;CLI Status Command Switch Options</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#cli-sessions-command-switch-subcommands">1.3.11.&nbsp;CLI Sessions Command Switch Subcommands</a></div>
                              <ul>
                                 <li>
                                    <div class="section-link"><a href="#cli-sessions-list-command-switch-subcommands">1.3.11.1.&nbsp;CLI Sessions List Command Switch Options</a></div>
                                 </li>
                              </ul>
                           </li>
                        </ul>
                     </li>
                     <li>
                        <div class="section-link"><a href="#example-single-command-lines">1.4.&nbsp;Example Single Command Lines</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#example-interactive-cli-command-sequences">1.5.&nbsp;Example Interactive CLI Command Sequences</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#example-nsys-stats">1.6.&nbsp;Example Stats Command Sequences</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#example-output-stats-option">1.7.&nbsp;Example Output from --stats Option</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#importing-qdstrm-files">1.8.&nbsp;Importing and Viewing Command Line Results Files</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#cli-analyze-mpi-codes">1.9.&nbsp;Using the CLI to Analyze MPI Codes</a></div>
                        <ul>
                           <li>
                              <div class="section-link"><a href="#cli-tracing-mpi-api-calls">1.9.1.&nbsp;Tracing MPI API calls</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#cli-profiling-apps-launched-mpirun">1.9.2.&nbsp;Using the CLI to Profile Applications Launched with mpirun</a></div>
                           </li>
                        </ul>
                     </li>
                  </ul>
               </li>
               <li>
                  <div class="section-link"><a href="#gui-profiling">2.&nbsp;Profiling from the GUI</a></div>
                  <ul>
                     <li>
                        <div class="section-link"><a href="#profiling-linux">2.1.&nbsp;Profiling Linux Targets from the GUI</a></div>
                        <ul>
                           <li>
                              <div class="section-link"><a href="#linux-connect-target">2.1.1.&nbsp;Connecting to the Target Device</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#linux-system-wide-profiling-options">2.1.2.&nbsp;System-Wide Profiling Options</a></div>
                              <ul>
                                 <li>
                                    <div class="section-link"><a href="#linux-x86-64">2.1.2.1.&nbsp;Linux x86_64</a></div>
                                 </li>
                                 <li>
                                    <div class="section-link"><a href="#linux-for-tegra">2.1.2.2.&nbsp;Linux for Tegra</a></div>
                                 </li>
                              </ul>
                           </li>
                           <li>
                              <div class="section-link"><a href="#linux-target-sampling-options">2.1.3.&nbsp;Target Sampling Options</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#linux-hotkey-trace-start-stop">2.1.4.&nbsp;Hotkey Trace Start/Stop</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#linux-launch-processes">2.1.5.&nbsp;Launching Processes</a></div>
                           </li>
                        </ul>
                     </li>
                     <li>
                        <div class="section-link"><a href="#profiling-windows">2.2.&nbsp;Profiling Windows Targets from the GUI</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#profiling-qnx">2.3.&nbsp;Profiling QNX Targets from the GUI</a></div>
                     </li>
                  </ul>
               </li>
               <li>
                  <div class="section-link"><a href="#exporter">3.&nbsp;Export Formats</a></div>
                  <ul>
                     <li>
                        <div class="section-link"><a href="#exporter-sqlite-schema">3.1.&nbsp;SQLite Schema Reference</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#exporter-sqlite-event-values">3.2.&nbsp;SQLite Schema Event Values</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#export-sqlite-examples">3.3.&nbsp;Common SQLite Examples</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#exporter-arrow">3.4.&nbsp;Arrow Format Description</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#exporter-json-txt">3.5.&nbsp;JSON and Text Format Description</a></div>
                     </li>
                  </ul>
               </li>
               <li>
                  <div class="section-link"><a href="#report-scripts">4.&nbsp;Report Scripts</a></div>
               </li>
               <li>
                  <div class="section-link"><a href="#migrate-nvprof">5.&nbsp;Migrating from NVIDIA nvprof</a></div>
               </li>
               <li>
                  <div class="section-link"><a href="#docker-profiling">6.&nbsp;Profiling in a Docker on Linux Devices</a></div>
               </li>
               <li>
                  <div class="section-link"><a href="#direct3d">7.&nbsp;Direct3D Trace</a></div>
                  <ul>
                     <li>
                        <div class="section-link"><a href="#d3d11-trace">7.1.&nbsp;D3D11 API trace</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#d3d12-trace">7.2.&nbsp;D3D12 API Trace</a></div>
                     </li>
                  </ul>
               </li>
               <li>
                  <div class="section-link"><a href="#wddm-queues">8.&nbsp;WDDM Queues</a></div>
               </li>
               <li>
                  <div class="section-link"><a href="#wddm-hw-scheduler">9.&nbsp;WDDM HW Scheduler</a></div>
               </li>
               <li>
                  <div class="section-link"><a href="#vulkan">10.&nbsp;Vulkan API Trace</a></div>
                  <ul>
                     <li>
                        <div class="section-link"><a href="#vulkan-overview">10.1.&nbsp;Vulkan Overview</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#vulkan-pipeline-creation">10.2.&nbsp;Pipeline Creation Feedback</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#vulkan-gpu-trace">10.3.&nbsp;Vulkan GPU Trace Notes</a></div>
                     </li>
                  </ul>
               </li>
               <li>
                  <div class="section-link"><a href="#stutter-analysis">11.&nbsp;Stutter Analysis</a></div>
                  <ul>
                     <li>
                        <div class="section-link"><a href="#fps-overview">11.1.&nbsp;FPS Overview</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#frame-health">11.2.&nbsp;Frame Health</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#gpu-memory-utilization">11.3.&nbsp;GPU Memory Utilization</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#vertical-sync">11.4.&nbsp;Vertical Synchronization</a></div>
                     </li>
                  </ul>
               </li>
               <li>
                  <div class="section-link"><a href="#openmp-trace">12.&nbsp;OpenMP Trace</a></div>
               </li>
               <li>
                  <div class="section-link"><a href="#os-runtime-trace">13.&nbsp;OS Runtime Libraries Trace</a></div>
                  <ul>
                     <li>
                        <div class="section-link"><a href="#osrt-locking-resource">13.1.&nbsp;Locking a Resource</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#osrt-limitations">13.2.&nbsp;Limitations</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#osrt-filters">13.3.&nbsp;OS Runtime Libraries Trace Filters</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#osrt-default-function">13.4.&nbsp;OS Runtime Default Function List</a></div>
                     </li>
                  </ul>
               </li>
               <li>
                  <div class="section-link"><a href="#nvtx-trace">14.&nbsp;NVTX Trace</a></div>
               </li>
               <li>
                  <div class="section-link"><a href="#cuda-trace">15.&nbsp;CUDA Trace</a></div>
                  <ul>
                     <li>
                        <div class="section-link"><a href="#cuda-gpu-memory-graph">15.1.&nbsp;CUDA GPU Memory Allocation Graph</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#unified-memory-trace">15.2.&nbsp;Unified Memory Transfer Trace</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#cuda-graph">15.3.&nbsp;CUDA Graph Trace</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#cuda-default-cli">15.4.&nbsp;CUDA Default Function List for CLI</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#cudnn-function-cli">15.5.&nbsp;cuDNN Function List for X86 CLI</a></div>
                     </li>
                  </ul>
               </li>
               <li>
                  <div class="section-link"><a href="#openacc-trace">16.&nbsp;OpenACC Trace</a></div>
               </li>
               <li>
                  <div class="section-link"><a href="#opengl-trace">17.&nbsp;OpenGL Trace</a></div>
                  <ul>
                     <li>
                        <div class="section-link"><a href="#opengl-cli">17.1.&nbsp;OpenGL Trace Using Command Line</a></div>
                     </li>
                  </ul>
               </li>
               <li>
                  <div class="section-link"><a href="#custom-etw-trace">18.&nbsp;Custom ETW Trace</a></div>
               </li>
               <li>
                  <div class="section-link"><a href="#gpu-metric-sampling">19.&nbsp;GPU Metrics</a></div>
               </li>
               <li>
                  <div class="section-link"><a href="#nvvideo-trace">20.&nbsp;NVIDIA Video Codec SDK Trace</a></div>
                  <ul>
                     <li>
                        <div class="section-link"><a href="#nvenc-default-function">20.1.&nbsp;NV Encoder API Functions Traced by Default</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#nvdec-default-function">20.2.&nbsp;NV Decoder API Functions Traced by Default</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#nvjpeg-default-function">20.3.&nbsp;NV JPEG API Functions Traced by Default</a></div>
                     </li>
                  </ul>
               </li>
               <li>
                  <div class="section-link"><a href="#communication-trace">21.&nbsp;Network Communication Profiling</a></div>
                  <ul>
                     <li>
                        <div class="section-link"><a href="#mpi-trace">21.1.&nbsp;MPI API Trace</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#openshmem-trace">21.2.&nbsp;OpenSHMEM Library Trace</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#ucx-trace">21.3.&nbsp;UCX Library Trace</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#nv-communications">21.4.&nbsp;NVIDIA NVSHMEM and NCCL Trace</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#NIC-sampling">21.5.&nbsp;NIC Metric Sampling</a></div>
                     </li>
                  </ul>
               </li>
               <li>
                  <div class="section-link"><a href="#report">22.&nbsp;Reading Your Report in GUI</a></div>
                  <ul>
                     <li>
                        <div class="section-link"><a href="#generate-new-report">22.1.&nbsp;Generating a New Report</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#open-report">22.2.&nbsp;Opening an Existing Report</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#report-sharing">22.3.&nbsp;Sharing a Report File</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#report-tab">22.4.&nbsp;Report Tab</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#analysis-summary-view">22.5.&nbsp;Analysis Summary View</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#timeline-view">22.6.&nbsp;Timeline View</a></div>
                        <ul>
                           <li>
                              <div class="section-link"><a href="#timeline">22.6.1.&nbsp;Timeline</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#events-view">22.6.2.&nbsp;Events View</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#function-table-modes">22.6.3.&nbsp;Function Table Modes</a></div>
                           </li>
                           <li>
                              <div class="section-link"><a href="#filter-dialog">22.6.4.&nbsp;Filter Dialog</a></div>
                           </li>
                        </ul>
                     </li>
                     <li>
                        <div class="section-link"><a href="#diagnostics-view">22.7.&nbsp;Diagnostics Summary View</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#symbol-resolution-view">22.8.&nbsp;Symbol Resolution Logs View</a></div>
                     </li>
                  </ul>
               </li>
               <li>
                  <div class="section-link"><a href="#multi-report-support">23.&nbsp;Adding Report to the Timeline</a></div>
                  <ul>
                     <li>
                        <div class="section-link"><a href="#multi-report-time-sync">23.1.&nbsp;Time Synchronization</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#multi-report-hierarchy">23.2.&nbsp;Timeline Hierarchy</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#multi-report-mpi-example">23.3.&nbsp;Example: MPI</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#multi-report-limitations">23.4.&nbsp;Limitations</a></div>
                     </li>
                  </ul>
               </li>
               <li>
                  <div class="section-link"><a href="#expert-systems">24.&nbsp;Using Nsight Systems Expert System</a></div>
               </li>
               <li>
                  <div class="section-link"><a href="#import-nvtxt">25.&nbsp;Import NVTXT</a></div>
               </li>
               <li>
                  <div class="section-link"><a href="#ballista-overview">26.&nbsp;Visual Studio Integration</a></div>
               </li>
               <li>
                  <div class="section-link"><a href="#troubleshooting">27.&nbsp;Troubleshooting</a></div>
                  <ul>
                     <li>
                        <div class="section-link"><a href="#general-troubleshooting">27.1.&nbsp;General Troubleshooting</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#cli-troubleshooting">27.2.&nbsp;CLI Troubleshooting</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#launch-stopped-linux">27.3.&nbsp;Launch Processes in Stopped State</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#gui-troubleshooting">27.4.&nbsp;GUI Troubleshooting</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#symbol-resolution-troubleshooting">27.5.&nbsp;Symbol Resolution</a></div>
                     </li>
                     <li>
                        <div class="section-link"><a href="#logging-troubleshooting">27.6.&nbsp;Logging</a></div>
                     </li>
                  </ul>
               </li>
               <li>
                  <div class="section-link"><a href="#other_resources">28.&nbsp;Other Resources</a></div>
               </li>
            </ul>
         </nav>
         <div id="resize-nav"></div>
         <nav id="search-results">
            <h2>Search Results</h2>
            <ol></ol>
         </nav>
         
         <div id="contents-container">
            <div id="breadcrumbs-container">
               <div id="release-info">User Guide
                  (<a href="../pdf/UserGuide.pdf">PDF</a>)
                  
                  -
                  
                  v2022.4.1
                  (<a href="https://developer.nvidia.com/nsight-systems">older</a>)
                  -
                  Last updated November 14, 2022
                  -
                  <a href="mailto:devtools-support@nvidia.com?subject=Nsight Systems Documentation Feedback: User Guide">Send Feedback</a>
                  -
                  <span class="st_facebook"></span><span class="st_twitter"></span><span class="st_linkedin"></span><span class="st_reddit"></span><span class="st_slashdot"></span><span class="st_tumblr"></span><span class="st_sharethis"></span></div>
            </div>
            <article id="contents">
               <div class="topic nested0" id="abstract"><a name="abstract" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#abstract" name="abstract" shape="rect">User Guide</a></h2>
                  <div class="body conbody">
                     <p class="p">
                        NVIDIA Nsight Systems user guide.
                        
                     </p>
                  </div>
               </div>
               <div class="topic nested0" id="cli-profiling"><a name="cli-profiling" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#cli-profiling" name="cli-profiling" shape="rect">1.&nbsp;Profiling from the CLI</a></h2>
                  <div class="body">
                     <p class="p"></p>
                  </div>
                  <div class="topic nested1" id="cli-installing"><a name="cli-installing" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#cli-installing" name="cli-installing" shape="rect">1.1.&nbsp;Installing the CLI on Your Target</a></h3>
                     <div class="body">
                        <p class="p">The <span class="ph">Nsight Systems</span> CLI provides a simple interface to collect
                           on a target without using the GUI. The collected data can then be copied to
                           any system and analyzed later.
                        </p>
                        <p class="p">The CLI is distributed in the Target directory of the standard
                           <span class="ph">Nsight Systems</span> download package. Users who want to install
                           the CLI as a standalone tool can do so by copying the files within the Target
                           directory. If you want the CLI output file (.qdstrm) to be auto-converted
                           (to .nsys-rep) after the analysis is complete, you will need to copy the host
                           directory as well.
                        </p>
                        <p class="p">If you wish to run the CLI without root (recommended mode), you will
                           want to install in a directory where you have full access.
                        </p>
                        <p class="p">Note that you must run the CLI on Windows as administrator.</p>
                     </div>
                  </div>
                  <div class="topic nested1" id="cli-options"><a name="cli-options" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#cli-options" name="cli-options" shape="rect">1.2.&nbsp;Command Line Options</a></h3>
                     <div class="body">
                        <p class="p">The <span class="ph">Nsight Systems</span> command lines can have one of two forms:
                        </p><pre xml:space="preserve"><span class="ph">nsys</span> [global_option]</pre><p class="p">or</p><pre xml:space="preserve"><span class="ph">nsys</span> [command_switch][optional command_switch_options][application] [optional application_options]</pre><p class="p">All command line options are case sensitive. For command switch options,
                           when short options are used, the parameters should follow the switch after a
                           space; e.g. <samp class="ph codeph">-s process-tree</samp>. When long options are used, the switch should be
                           followed by an equal sign and then the parameter(s); e.g. <samp class="ph codeph">--sample=process-tree</samp>.
                        </p>
                        <p class="p">For this version of <span class="ph">Nsight Systems</span>, if you launch a process
                           from the command line to begin analysis, the launched process will be terminated
                           when collection is complete, including runs with --duration set, unless the
                           user specifies the --kill none option (details below). The exception is that if the
                           user uses NVTX, cudaProfilerStart/Stop, or hotkeys to control the duration,
                           the application will continue unless --kill is set.
                        </p>
                        <p class="p">The <span class="ph">Nsight Systems</span> CLI supports concurrent analysis by using
                           sessions. Each <span class="ph">Nsight Systems</span> session is defined by a sequence
                           of CLI commands that define one or more collections (e.g. when and what data
                           is collected). A session begins with either a start, launch, or profile
                           command. A session ends with a shutdown command, when a profile command
                           terminates, or, if requested, when all the process tree(s) launched in the
                           session exit. Multiple sessions can run concurrently on the same system.
                        </p>
                     </div>
                     <div class="topic nested2" id="cli-global-options"><a name="cli-global-options" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cli-global-options" name="cli-global-options" shape="rect">1.2.1.&nbsp;CLI Global Options</a></h3>
                        <div class="body">
                           <div class="p">
                              <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                                 <tr class="sthead">
                                    <th valign="bottom" align="left" id="d54e162" class="stentry" width="6.666666666666667%" rowspan="1" colspan="1">Short</th>
                                    <th valign="bottom" align="left" id="d54e165" class="stentry" width="26.666666666666668%" rowspan="1" colspan="1">Long</th>
                                    <th valign="bottom" align="left" id="d54e168" class="stentry" width="66.66666666666667%" rowspan="1" colspan="1">Description</th>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e162" class="stentry" rowspan="1" colspan="1">-h</td>
                                    <td valign="top" headers="d54e165" class="stentry" rowspan="1" colspan="1">--help</td>
                                    <td valign="top" headers="d54e168" class="stentry" rowspan="1" colspan="1">Help message providing information about available command
                                       switches and their options.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e162" class="stentry" rowspan="1" colspan="1">-v</td>
                                    <td valign="top" headers="d54e165" class="stentry" rowspan="1" colspan="1">--version</td>
                                    <td valign="top" headers="d54e168" class="stentry" rowspan="1" colspan="1">Output <span class="ph">Nsight Systems</span> CLI version information.
                                    </td>
                                 </tr>
                              </table>
                           </div>
                        </div>
                     </div>
                  </div>
                  <div class="topic nested1" id="cli-command-switches"><a name="cli-command-switches" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#cli-command-switches" name="cli-command-switches" shape="rect">1.3.&nbsp;CLI Command Switches</a></h3>
                     <div class="body">
                        <p class="p">The <span class="ph">Nsight Systems</span> command line interface can be used in two
                           modes. You may launch your application and begin analysis with options
                           specified to the <samp class="ph codeph"><span class="ph">nsys</span> profile</samp> command.
                           Alternatively, you can control the launch of an application and data
                           collection using interactive CLI commands.
                        </p>
                        <div class="p">
                           <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                              <tr class="sthead">
                                 <th valign="bottom" align="left" id="d54e228" class="stentry" width="20%" rowspan="1" colspan="1">Command</th>
                                 <th valign="bottom" align="left" id="d54e231" class="stentry" width="80%" rowspan="1" colspan="1">Description</th>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e228" class="stentry" rowspan="1" colspan="1">profile</td>
                                 <td valign="top" headers="d54e231" class="stentry" rowspan="1" colspan="1">A fully formed profiling description requiring and accepting
                                    no further input. The command switch options used (see below table)
                                    determine when the collection starts, stops, what collectors are used
                                    (e.g. API trace, IP sampling, etc.), what processes are monitored, etc.
                                    
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e228" class="stentry" rowspan="1" colspan="1">start</td>
                                 <td valign="top" headers="d54e231" class="stentry" rowspan="1" colspan="1">Start a collection in interactive mode. The start command can
                                    be executed before or after a launch command.
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e228" class="stentry" rowspan="1" colspan="1">stop</td>
                                 <td valign="top" headers="d54e231" class="stentry" rowspan="1" colspan="1">Stop a collection that was started in interactive mode. When
                                    executed, all active collections stop, the CLI process terminates but the
                                    application continues running.
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e228" class="stentry" rowspan="1" colspan="1">cancel</td>
                                 <td valign="top" headers="d54e231" class="stentry" rowspan="1" colspan="1">Cancels an existing collection started in interactive mode.
                                    All data already collected in the current collection is discarded.
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e228" class="stentry" rowspan="1" colspan="1">launch</td>
                                 <td valign="top" headers="d54e231" class="stentry" rowspan="1" colspan="1">In interactive mode, launches an application in an environment
                                    that supports the requested options. The launch command can be executed
                                    before or after a start command.
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e228" class="stentry" rowspan="1" colspan="1">shutdown</td>
                                 <td valign="top" headers="d54e231" class="stentry" rowspan="1" colspan="1">Disconnects the CLI process from the launched application and
                                    forces the CLI process to exit. If a collection is pending or active,
                                    it is cancelled
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e228" class="stentry" rowspan="1" colspan="1">export</td>
                                 <td valign="top" headers="d54e231" class="stentry" rowspan="1" colspan="1">Generates an export file from an existing .nsys-rep file. For more
                                    information about the exported formats see the /documentation/nsys-exporter
                                    directory in your <span class="ph">Nsight Systems</span> installation directory.
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e228" class="stentry" rowspan="1" colspan="1">stats</td>
                                 <td valign="top" headers="d54e231" class="stentry" rowspan="1" colspan="1">Post process existing <span class="ph">Nsight Systems</span> result,
                                    either in .nsys-rep or SQLite format, to generate statistical information.
                                    
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e228" class="stentry" rowspan="1" colspan="1">analyze</td>
                                 <td valign="top" headers="d54e231" class="stentry" rowspan="1" colspan="1">Post process existing <span class="ph">Nsight Systems</span> result,
                                    either in .nsys-rep or SQLite format, to generate expert systems report.
                                    
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e228" class="stentry" rowspan="1" colspan="1">status</td>
                                 <td valign="top" headers="d54e231" class="stentry" rowspan="1" colspan="1">Reports on the status of a CLI-based collection or the
                                    suitability of the profiling environment.
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e228" class="stentry" rowspan="1" colspan="1">sessions</td>
                                 <td valign="top" headers="d54e231" class="stentry" rowspan="1" colspan="1">Gives information about all sessions running on the system.</td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e228" class="stentry" rowspan="1" colspan="1">nvprof</td>
                                 <td valign="top" headers="d54e231" class="stentry" rowspan="1" colspan="1">Special option to help with transition from legacy NVIDIA
                                    nvprof tool. Calling <samp class="ph codeph"><span class="ph">nsys</span> nvprof [options]</samp>
                                    will provide the best available translation of <samp class="ph codeph">nvprof [options]</samp>
                                    See <strong class="ph b">Migrating from NVIDIA nvprof</strong> topic for details. No additional
                                    functionality of <span class="ph">nsys</span> will be available when using
                                    this option. Note: Not available on IBM Power targets.
                                 </td>
                              </tr>
                           </table>
                        </div>
                     </div>
                     <div class="topic nested2" id="cli-profile-command-switch-options"><a name="cli-profile-command-switch-options" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cli-profile-command-switch-options" name="cli-profile-command-switch-options" shape="rect">1.3.1.&nbsp;CLI Profile Command Switch Options</a></h3>
                        <div class="body">
                           <p class="p">After choosing the <samp class="ph codeph">profile</samp> command switch, the
                              following options are available. Usage:
                           </p><pre xml:space="preserve"><span class="ph">nsys</span> [global-options] profile [options] &lt;application&gt; [application-arguments]</pre><div class="p">
                              <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                                 <tr class="sthead">
                                    <th valign="bottom" align="left" id="d54e396" class="stentry" width="8.695652173913043%" rowspan="1" colspan="1">Short</th>
                                    <th valign="bottom" align="left" id="d54e399" class="stentry" width="13.043478260869565%" rowspan="1" colspan="1">Long</th>
                                    <th valign="bottom" align="left" id="d54e402" class="stentry" width="8.695652173913043%" rowspan="1" colspan="1">Possible Parameters</th>
                                    <th valign="bottom" align="left" id="d54e405" class="stentry" width="8.695652173913043%" rowspan="1" colspan="1">Default</th>
                                    <th valign="bottom" align="left" id="d54e408" class="stentry" width="60.8695652173913%" rowspan="1" colspan="1">Switch Description</th>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--accelerator-trace</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">none,nvmedia</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Collect other accelerators workload trace from the hardware
                                       engine units. Only available on <span class="ph">Nsight Systems Embedded Platforms Edition</span>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">-c</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--capture-range</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">none, cudaProfilerApi, hotkey, nvtx</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">When --capture-range is used, profiling will
                                       start only when appropriate start API or hotkey is invoked. If
                                       --capture-range is set to none, start/stop API calls and hotkeys will
                                       be ignored. Note: Hotkey works for graphic applications only.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">-d</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--duration</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">&lt; seconds &gt;</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">NA</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Collection duration in seconds, duration must be greater than
                                       zero. Note that the profiler does not detach from the application, it
                                       lives until application termination.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">-e</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--env-var</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">A=B</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">NA</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Set environment variable(s) for the application process to be
                                       launched. Environment variables should be defined as A=B. Multiple
                                       environment variables can be specified as A=B,C=D.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--export</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">arrow, hdf, json, sqlite, text, none</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Create additional output file(s) based on the data collected.
                                       This option can be given more than once.
                                       WARNING: If the collection captures a large amount of data, creating
                                       the export file may take several minutes to complete.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">-f</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--force-overwrite</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">If true, overwrite all existing result files with same output
                                       filename (.qdstrm, .nsys-rep, .arrows, .h5, .json, .sqlite, .txt).
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--help</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">&lt;tag&gt;</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Print the help message. The option can take one optional
                                       argument that will be used as a tag. If a tag is provided, only options
                                       relevant to the tag will be printed.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--kill</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">none, sigkill, sigterm, signal number</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">sigterm</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Send signal to the target application's process group. Can be
                                       used with --duration or range markers.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">-n</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--inherit-environment</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">When true, the current environment variables and the tool’s
                                       environment variables will be specified for the launched process. When
                                       false, only the tool’s environment variables will be specified for the
                                       launched process.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">-o</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--output</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">&lt; filename &gt;</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">report#</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Set report file name. Any %q{ENV_VAR} pattern in the filename
                                       will be substituted with the value of the environment variable. Any %h
                                       pattern in the filename will be substituted with the hostname of the
                                       system. Any %p pattern in the filename will be substituted with the
                                       PID of the target process or the PID of the root process if there is a
                                       process tree. Any %% pattern in the filename will be substituted with %.
                                       Default is report#.{qdstrm,nsys-rep,sqlite,h5,txt,arrows,json} in the working directory.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--session-new</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">[a-Z][0-9,a-Z,spaces]</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">profile-&lt;id&gt;-&lt;application&gt;</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Name the session created by the command. Name must start with
                                       an alphabetical character followed by printable or space characters.
                                       Any <samp class="ph codeph">%q{ENV_VAR}</samp> pattern will be substituted with the
                                       value of the environment variable. Any <samp class="ph codeph">%h</samp> pattern
                                       will be substituted with the hostname of the system. Any
                                       <samp class="ph codeph">%%</samp> pattern will be substituted with <samp class="ph codeph">%</samp>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">-w</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--show-output</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">If true, send target process’ stdout and stderr streams to the
                                       console and stdout/stderr files which are added to the QDSTRM file.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--stats</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Generate summary statistics after the collection. WARNING:
                                       When set to true, an SQLite database will be created after the
                                       collection. If the collection captures a large amount of data, creating
                                       the database file may take several minutes to complete.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">-t</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--trace</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">cuda, nvtx, cublas, cublas-verbose, cusparse, cusparse-verbose,
                                       cudnn, opengl, opengl-annotations, openacc, openmp, osrt,
                                       mpi, nvvideo, vulkan, vulkan-annotations, dx11, dx11-annotations, dx12,
                                       dx12-annotations, oshmem, ucx, wddm, nvmedia, none
                                    </td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">cuda, opengl, nvtx, osrt</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Select the API(s) to be traced. The osrt switch controls the
                                       OS runtime libraries tracing. Multiple APIs can be selected, separated
                                       by commas only (no spaces). Since OpenACC, cuDNN and cuBLAS APIs are
                                       tightly linked with CUDA, selecting one of those APIs will automatically
                                       enable CUDA tracing. Reflex SDK latency markers will be automatically
                                       collected when DX or vulkan API trace is enabled. See information on
                                       --mpi-impl option below if mpi is selected. If '&lt;api&gt;-annotations'
                                       is selected, the corresponding API will also be traced. If the none
                                       option is selected, no APIs are traced and no other API can be selected.
                                       Note: cublas, cudnn, nvvideo, opengl, and vulkan are
                                       not available on IBM Power target.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">-x</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--stop-on-exit</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">If true, stop collecting automatically when the launched
                                       process has exited or when the duration expires - whichever occurs
                                       first.  If false, duration must be set and the collection stops only
                                       when the duration expires. <span class="ph">Nsight Systems</span> does not
                                       officially support runs longer than 5 minutes.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">-y</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--delay</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">&lt; seconds &gt;</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">0</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Collection start delay in seconds.</td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--dx-force-declare-adapter-removal-support</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">The <span class="ph">Nsight Systems</span> trace initialization involves creating a D3D
                                       device and discarding it. Enabling this flag makes a call to
                                       DXGIDeclareAdapterRemovalSupport() before device creation. Requires DX11 or DX12 trace
                                       to be enabled.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--dx12-gpu-workload</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false, individual, batch, none</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">individual</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">If individual or true, trace each DX12 workload's GPU activity individually.
                                       If batch, trace DX12 workloads' GPU activity in ExecuteCommandLists call batches.
                                       If none or false, do not trace DX12 workloads' GPU activity.
                                       Note that this switch is applicable only when --trace=dx12 is specified.
                                       This option is only supported on Windows targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--dx12-wait-calls</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">If true, trace wait calls that block on fences for DX12. Note
                                       that this switch is applicable only when --trace=dx12 is specified. This
                                       option is only supported on Windows targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--gpu-metrics-device</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">GPU ID, help, all, none</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Collect GPU Metrics from specified devices.
                                       Determine GPU IDs by using <samp class="ph codeph">--gpu-metrics-device=help</samp>
                                       switch.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--gpu-metrics-frequency</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">integer</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">10000</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Specify GPU Metrics sampling frequency. Minimum supported
                                       frequency is 10 (Hz). Maximum supported frequency is 200000 (Hz).
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--gpu-metrics-set</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">index, alias</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Specify metric set for GPU Metrics. The argument must be one
                                       of indices or aliases reported by<samp class="ph codeph">--gpu-metrics-set=help</samp>
                                       switch. If not specified, the default is the first metric set that
                                       supports all selected GPUs.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--gpuctxsw</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true,false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Trace GPU context switches. Note that this requires driver
                                       r435.17 or later and root permission. Not supported on IBM Power
                                       targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--opengl-gpu-workload</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">If true, trace the OpenGL workloads' GPU activity. Note that
                                       this switch is applicable only when --trace=opengl is specified. This
                                       option is not supported on IBM Power targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--vulkan-gpu-workload</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false, individual, batch, none</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">individual</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">If individual or true, trace each Vulkan workload's GPU activity individually.
                                       If batch, trace Vulkan workloads' GPU activity in vkQueueSubmit call batches.
                                       If none or false, do not trace Vulkan workloads' GPU activity.
                                       Note that this switch is applicable only when --trace=vulkan is specified. This option
                                       is not supported on QNX.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--wddm-additional-events</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">If true, collect additional range of ETW events, including
                                       context status, allocations, sync wait and signal events, etc. Note
                                       that this switch is applicable only when --trace=wddm is specified.
                                       This option is only supported on Windows targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--accelerator-trace</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">none, nvmedia</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Collect other accelerators workload trace from the hardware
                                       engine units. Available in <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">-b</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--backtrace</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">auto,fp,lbr,dwarf,none</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Select the backtrace method to use while sampling. The option
                                       'lbr' uses Intel(c) Corporation's Last Branch Record registers, available only
                                       with Intel(c) CPUs codenamed Haswell and later. The option 'fp' is frame
                                       pointer and assumes that frame pointers were enabled during compilation.
                                       The option 'dwarf' uses DWARF's CFI (Call Frame Information). Setting the
                                       value to 'none' can reduce collection overhead.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--clock-frequency-changes</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Collect clock frequency changes. Available only in <span class="ph">Nsight Systems Embedded Platforms Edition</span>
                                       and Arm server (SBSA) platforms
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--cpu-cluster-events</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">0x16, 0x17, ..., none</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Collect per-cluster Uncore PMU counters. Multiple values can
                                       be selected, separated by commas only (no spaces). Use the
                                       --cpu-cluster-events=help switch to see the full list of values.
                                       Available in <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--cpu-core-events (<span class="ph">Nsight Systems Embedded Platforms Edition</span>)
                                    </td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">0x11,0x13,...,none</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Collect per-core PMU counters. Multiple values can be
                                       selected, separated by commas only (no spaces). Use the
                                       <samp class="ph codeph">--cpu-core-events=help</samp> switch to see the full list of
                                       values.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--cpu-core-events (not <span class="ph">Nsight Systems Embedded Platforms Edition</span>)
                                    </td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">'help' or the end users selected events in the format 'x,y'</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">'2' i.e. Instructions Retired</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Select the CPU Core events to sample.  Use the
                                       <samp class="ph codeph">--cpu-core-events=help</samp> switch to see the full list of
                                       events and the number of events that can be collected simultaneously.
                                       Multiple values can be selected, separated by commas only (no spaces).
                                       Use the --event-sample switch to enable.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--event-sampling-frequency</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">Integers from 1 to 20 Hz</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">3</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">The sampling frequency used to collect event counts. Minimum
                                       event sampling frequency is 1 Hz. Maximum event sampling frequency is
                                       20 Hz. Not available in <span class="ph">Nsight Systems Embedded Platforms Edition</span>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--cpu-socket-events</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">0x2a, 0x2c, ..., none</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Collect per-socket Uncore PMU counters. Multiple values can
                                       be selected, separated by commas only (no spaces). Use the
                                       --cpu-socket-events=help switch to see the full list of values.
                                       Available in <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--cpuctxsw</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">process-tree, system-wide, none</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">process-tree</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Trace OS thread scheduling activity. Select 'none' to disable
                                       tracing CPU context switches. Depending on the platform, some values may
                                       require admin or root privileges. Note: if the --sample switch is set to a value
                                       other than 'none', the --cpuctxsw setting is hardcoded to the same value as the
                                       --sample switch. If --sample=none and a target application is launched, the
                                       default is 'process-tree', otherwise the default is 'none'. Requires
                                       <samp class="ph codeph">--sampling-trigger=perf</samp> switch in
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span></td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--os-events</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">'help' or the end users selected events in the format 'x,y'</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Select the OS events to sample. Use the --os-events=help
                                       switch to see the full list of events. Multiple values can be selected,
                                       separated by commas only (no spaces). Use the --event-sample switch to
                                       enable. Not available on <span class="ph">Nsight Systems Embedded Platforms Edition</span>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--event-sample</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">system-wide, none</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Use the --cpu-core-events=help and the --os-events=help
                                       switches to see the full list of events. If event sampling is enabled
                                       and no events are selected, the CPU Core event 'Instructions Retired' is
                                       selected by default. Not available on <span class="ph">Nsight Systems Embedded Platforms Edition</span>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--nic-metrics</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Collect metrics from supported NIC/HCA devices. Not available
                                       on <span class="ph">Nsight Systems Embedded Platforms Edition</span>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--osrt-backtrace-depth</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">integer</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">24</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Set the depth for the backtraces collected for OS runtime
                                       libraries calls.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--osrt-backtrace-stack-size</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">integer</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">6144</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Set the stack dump size, in bytes, to generate
                                       backtraces for OS runtime libraries calls.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--osrt-backtrace-threshold</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">nanoseconds</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">80000</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Set the duration, in nanoseconds, that all OS runtime
                                       libraries calls must execute before backtraces are collected. 
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--osrt-threshold</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">&lt; nanoseconds &gt;</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">1000 ns</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Set the duration, in nanoseconds, that Operating System Runtime (osrt) APIs must
                                       execute before they are traced. Values significantly less than 1000 may cause
                                       significant overhead and result in extremely large result files. Note:
                                       Not available for IBM Power targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--qnx-kernel-events</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">class/event,event,class/event:mode,class:mode,help,none</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Multiple values can be selected, separated by commas only (no spaces). See
                                       the --qnx-kernel-events-mode switch description for ':mode' format. Use the
                                       '--qnx-kernel-events=help' switch to see the full list of values.
                                       Example: '--qnx-kernel-events=8/1:system:wide,_NTO_TRACE_THREAD:process:fast,
                                       _NTO_TRACE_KERCALLENTER/__KER_BAD,_NTO_TRACE_COMM,13'.
                                       Collect QNX kernel events.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--qnx-kernel-events-mode</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">system,process,fast,wide</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">system:fast</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Values are separated by a colon (':') only (no spaces). 'system' and 'process'
                                       cannot be specified at the same time. 'fast' and 'wide' cannot be specified at the same
                                       time. Please check the QNX documentation to determine when to select the 'fast' or 'wide'
                                       mode. Specify the default mode for QNX kernel events collection.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--resolve-symbols</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true,false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Resolve symbols of captured samples and backtraces.</td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--injection-use-detours</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true,false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Use detours for injection. If false, process injection will be performed by 
                                       windows hooks which allows to bypass anti-cheat software.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">-s</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--sample</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">process-tree, system-wide, none</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">process-tree</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Select how to collect CPU IP/backtrace samples. If 'none' is
                                       selected, CPU sampling is disabled. Depending on the platform, some values may
                                       require admin or root privileges. If a target application is launched, the
                                       default is 'process-tree', otherwise, the default is 'none'. Note: 'system-wide'
                                       is not available on all platforms. Note: If set to 'none', CPU context switch
                                       data will still be collected unless the --cpuctxsw switch is set to 'none'.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--samples-per-backtrace</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">integer &lt;= 32</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">1</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">The number of CPU IP samples collected for every CPU IP/backtrace sample
                                       collected. For example, if set to 4, on the fourth CPU IP sample collected, a
                                       backtrace will also be collected. Lower values increase the amount of data collected.
                                       Higher values can reduce collection overhead and reduce the number of CPU IP
                                       samples dropped. If DWARF backtraces are collected, the default is 4, otherwise the
                                       default is 1. This option is not available on <span class="ph">Nsight Systems Embedded Platforms Edition</span>
                                       or on non-Linux targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--sampling-frequency</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">100 &lt; integers &lt; 8000</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">1000</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Specify the sampling/backtracing frequency. The minimum
                                       supported frequency is 100 Hz. The maximum supported frequency is 8000
                                       Hz. This option is supported only on QNX, Linux for Tegra, and Windows
                                       targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--sampling-period (<span class="ph">Nsight Systems Embedded Platforms Edition</span>)
                                    </td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">integer</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">determined dynamically</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">The number of CPU Cycle events counted before a CPU
                                       instruction pointer (IP) sample is collected. If configured,
                                       backtraces may also be collected. The smaller the sampling period,
                                       the higher the sampling rate. Note that smaller sampling periods will
                                       increase overhead and significantly increase the size of the result
                                       file(s). Requires <samp class="ph codeph">--sampling-trigger=perf</samp> switch.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--sampling-period (not <span class="ph">Nsight Systems Embedded Platforms Edition</span>)
                                    </td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">integer</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">determined dynamically</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">The number of events counted before a CPU instruction pointer
                                       (IP) sample is collected. The event used to trigger the collection of
                                       a sample is determined dynamically. For example, on Intel based
                                       platforms, it will probably be "Reference Cycles" and on AMD platforms,
                                       "CPU Cycles". If configured, backtraces may also be collected. The
                                       smaller the sampling period, the higher the sampling rate. Note that
                                       smaller sampling periods will increase overhead and significantly
                                       increase the size of the result file(s). This option is available only
                                       on Linux targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--sampling-trigger</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">timer, sched, perf, cuda</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">timer,sched</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Specify backtrace collection trigger. Multiple APIs can be
                                       selected, separated by commas only (no spaces). Available on
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span> targets only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--el1-sampling</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Enable EL1 sampling. Available in
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--el1-sampling-config</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">&lt; filepath config.json &gt;</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">EL1 sampling config. Available in
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span> only. 
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--xhv-trace</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">&lt; filepath pct.json &gt;</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Collect hypervisor trace. Available in
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--xhv-trace-events</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">all, none, core, sched, irq, trap</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">all</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Available in
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--etw-provider</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">"&lt;name&gt;,&lt;guid&gt;", or path to JSON file</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Add custom ETW trace provider(s). If you want to specify more
                                       attributes than Name and GUID, provide a JSON configuration file as
                                       as outlined below. This switch can be used multiple times to add
                                       multiple providers. Note: Only available for Windows targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--retain-etw-files</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Retain ETW files generated by the trace, merge and move the
                                       files to the output directory.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--cuda-flush-interval</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">milliseconds</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">See Description</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Set the interval, in milliseconds, when buffered CUDA data is automatically
                                       saved to storage. CUDA data buffer saves may cause profiler overhead. Buffer save
                                       behavior can be controlled with this switch.
                                       
                                       If the CUDA flush interval is set to 0 on systems running CUDA 11.0 or newer, buffers are
                                       saved when they fill. If a flush interval is set to a non-zero value on such systems,
                                       buffers are saved only when the flush interval expires. If a flush interval is set and
                                       the profiler runs out of available buffers before the flush interval expires, additional
                                       buffers will be allocated as needed. In this case, setting a flush interval can reduce
                                       buffer save overhead but increase memory use by the profiler.
                                       
                                       If the flush interval is set to 0 on systems running older versions of CUDA,
                                       buffers are saved at the end of the collection. If the profiler runs out of available
                                       buffers, additional buffers are allocated as needed. If a flush interval is set to a
                                       non-zero value on such systems, buffers are saved when the flush interval expires.
                                       A cuCtxSynchronize call may be inserted into the workflow before the buffers
                                       are saved which will cause application overhead. In this case, setting a flush interval
                                       can reduce memory use by the profiler but may increase save overhead.
                                       
                                       For collections over 30 seconds an interval of 10 seconds is recommended.
                                       Default is 10000 for <span class="ph">Nsight Systems Embedded Platforms Edition</span> and 0 otherwise.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--cuda-memory-usage</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Track the GPU memory usage by CUDA kernels. Applicable only
                                       when CUDA tracing is enabled. Note: This feature may cause significant
                                       runtime overhead.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--cuda-um-cpu-page-faults</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">This switch tracks the page faults that occur
                                       when CPU code tries to access a memory page that resides on the device.
                                       Note that this feature may cause significant runtime overhead.
                                       Not available on <span class="ph">Nsight Systems Embedded Platforms Edition</span>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--cuda-um-gpu-page-faults</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">This switch tracks the page faults that occur
                                       when GPU code tries to access a memory page that resides on the host.
                                       Note that this feature may cause significant runtime overhead.
                                       Not availalbe on <span class="ph">Nsight Systems Embedded Platforms Edition</span>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--cudabacktrace</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">all, none, kernel, memory, sync, other</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">When tracing CUDA APIs, enable the collection of a backtrace
                                       when a CUDA API is invoked. Significant runtime overhead may occur.
                                       Values may be combined using ','. Each value except 'none' may be
                                       appended with a threshold after ':'. Threshold is duration, in
                                       nanoseconds, that CUDA APIs must execute before backtraces are
                                       collected, e.g. 'kernel:500'. Default value for each threshold is
                                       1000ns (1us). Note: CPU sampling must be enabled. Note: Not
                                       available on IBM Power targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--cuda-graph-trace</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">graph, node</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">graph</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">If 'graph' is selected, CUDA graphs will be traced as a whole
                                       and node activities will not be collected. This will reduce overhead
                                       to a minimum, but requires CUDA driver version 515.43 or higher.
                                       If 'node' is selected, node activities will be collected, but CUDA graphs
                                       will not be traced as a whole. This may cause significant runtime overhead.
                                       Default is 'graph' if available, otherwise default is 'node'.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--nvtx-domain-include</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">default, &lt;domain_names&gt;</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Choose to only include NVTX events from a comma separated
                                       list of domains. 'default' filters the NVTX default domain. A domain
                                       with this name or commas in a domain name must be escaped with '\'.
                                       Note: Only one of --nvtx-domain-include and --nvtx-domain-exclude can
                                       be used. This option is only applicable when --trace=nvtx is specified.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--nvtx-domain-exclude</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">default, &lt;domain_names&gt;</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Choose to exclude NVTX events from a comma separated list of
                                       domains. 'default' excludes NVTX events without a domain. A domain
                                       with this name or commas in a domain name must be escaped with '\'.
                                       Note: Only one of --nvtx-domain-include and --nvtx-domain-exclude can
                                       be used. This option is only applicable when --trace=nvtx is specified.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">-p</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--nvtx-capture</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">range@domain, range, range@*</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Specify NVTX range and domain to trigger the profiling session.
                                       This option is applicable only when used along with --capture-range=nvtx.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--auto-report-name</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Derive report file name from collected data uses details of profiled graphics
                                       application. Format: [Process Name][GPU Name][Window Resolution][Graphics API]
                                       Timestamp .nsys-rep If true, automatically generate report file
                                       names.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--capture-range-end</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">none, stop, stop-shutdown, repeat[:N], repeat-shutdown:N</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">stop-shutdown</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Specify the desired behavior when a capture range ends.
                                       Applicable only when used along with --capture-range option. If
                                       <samp class="ph codeph">none</samp>, capture range end will be ignored. If
                                       <samp class="ph codeph">stop</samp>, collection will stop at capture range end. Any
                                       subsequent capture ranges will be ignored. Target app will continue
                                       running. If <samp class="ph codeph">stop-shutdown</samp>, collection will stop at
                                       capture range end and session will be shutdown. If
                                       <samp class="ph codeph">repeat[:N]</samp>, collection will stop at capture range end
                                       and subsequent capture ranges will trigger more collections. Use the
                                       optional <samp class="ph codeph">:N</samp> to specify max number of capture ranges
                                       to be honored. Any subsequent capture ranges will be ignored once N
                                       capture ranges are collected. If <samp class="ph codeph">repeat-shutdown:N</samp>,
                                       same behavior as <samp class="ph codeph">repeat:N</samp> but session will be shutdown
                                       after N ranges. For <samp class="ph codeph">stop-shutdown</samp> and
                                       <samp class="ph codeph">repeat-shutdown:N</samp>, as always, use --kill option to
                                       specify whether target app should be terminated when shutting down
                                       session.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--command-file</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">&lt; filename &gt;</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Open a file that contains profile switches and parse the
                                       switches. Note additional switches on the command line will override
                                       switches in the file. This flag can be specified more than once.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--duration-frames</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">60 &lt;= integer</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Stop the recording session after this many frames have been captured.
                                       Note when it is selected cannot include any other stop options.
                                       If not specified, the default is disabled.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--ftrace</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Collect ftrace events. Argument should list events to
                                       collect as: subsystem1/event1,subsystem2/event2. Requires root. No
                                       ftrace events are collected by default. Note: Not available on IBM
                                       Power targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--ftrace-keep-user-config</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Skip initial ftrace setup and collect already configured
                                       events. Default resets the ftrace configuration.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--hotkey-capture</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">'F1' to 'F12'</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">'F12'</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Hotkey to trigger the profiling session. Note that this
                                       switch is applicable only when --capture-range=hotkey is specified.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--isr</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Trace Interrupt Service Routines (ISRs) and Deferred
                                       Procedure Calls (DPCs). Requires administrative privileges. Available
                                       only on Windows devices.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--mpi-impl</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">openmpi,mpich</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">openmpi</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">When using --trace=mpi to trace MPI APIs use --mpi-impl to
                                       specify which MPI implementation the application is using. If no MPI implementation is
                                       specified, <span class="ph">nsys</span> tries to automatically detect it based on the dynamic
                                       linker's search path. If this fails, 'openmpi' is used. Calling --mpi-impl without
                                       --trace=mpi is not supported.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--process-scope</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">main, process-tree, system-wide</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">main</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Select which process(es) to trace. Available in
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span> only. <span class="ph">Nsight Systems Workstation Edition</span> will
                                       always trace system-wide in this version of the tool.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--run-as</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">&lt; username &gt;</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Run the target application as the specified username. If not
                                       specified, the target application will be run by the same user as
                                       <span class="ph">Nsight Systems</span>. Requires root privileges. Available for
                                       Linux targets only. 
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--start-frame-index</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">1 &lt;= integer</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Start the recording session when the frame index reaches the frame number
                                       preceding the start frame index. Note when it is selected cannot include any other start
                                       options. If not specified, the default is disabled.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--trace-fork-before-exec</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">If true, trace any child process after fork and before they
                                       call one of the exec functions. Beware, tracing in this interval
                                       relies on undefined behavior and might cause your application to crash
                                       or deadlock. Note: This option is only available on Linux target platforms.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--vsync</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Collect vsync events. If collection of vsync events is
                                       enabled, display/display_scanline ftrace events will also be captured.
                                       Available in <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--wait</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">primary,all</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">all</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">If primary, the CLI will wait on the application process
                                       termination. If all, the CLI will additionally wait on re-parented
                                       processes created by the application.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--soc-metrics</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">true,false</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Collect SOC Metrics. Available in
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--soc-metrics-frequency</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">integer</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">100000</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Specify SOC Metrics sampling frequency. Minimum supported
                                       frequency is '100' (Hz). Maximum supported frequency is '1000000'
                                       (Hz). Available in <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e396" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e399" class="stentry" rowspan="1" colspan="1">--soc-metrics-set</td>
                                    <td valign="top" headers="d54e402" class="stentry" rowspan="1" colspan="1">see description</td>
                                    <td valign="top" headers="d54e405" class="stentry" rowspan="1" colspan="1">see description</td>
                                    <td valign="top" headers="d54e408" class="stentry" rowspan="1" colspan="1">Specify metric set for SOC Metrics sampling. The option
                                       argument must be one of indices or aliases reported by
                                       <samp class="ph codeph">--soc-metrics-set=help</samp> switch. Default is the first
                                       supported set. Available in <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                              </table>
                           </div>
                        </div>
                     </div>
                     <div class="topic nested2" id="cli-start-command-switch-options"><a name="cli-start-command-switch-options" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cli-start-command-switch-options" name="cli-start-command-switch-options" shape="rect">1.3.2.&nbsp;CLI Start Command Switch Options</a></h3>
                        <div class="body">
                           <p class="p">After choosing the <samp class="ph codeph">start</samp> command switch, the following
                              options are available. Usage:
                           </p><pre xml:space="preserve"><span class="ph">nsys</span> [global-options] start [options]</pre><div class="p">
                              <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                                 <tr class="sthead">
                                    <th valign="bottom" align="left" id="d54e2061" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Short</th>
                                    <th valign="bottom" align="left" id="d54e2064" class="stentry" width="14.285714285714285%" rowspan="1" colspan="1">Long</th>
                                    <th valign="bottom" align="left" id="d54e2067" class="stentry" width="19.047619047619047%" rowspan="1" colspan="1">Possible Parameters</th>
                                    <th valign="bottom" align="left" id="d54e2070" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Default</th>
                                    <th valign="bottom" align="left" id="d54e2073" class="stentry" width="47.61904761904762%" rowspan="1" colspan="1">Switch Description</th>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--accelerator-trace</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">none,nvmedia</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Collect other accelerators workload trace from the hardware
                                       engine units. Only available on <span class="ph">Nsight Systems Embedded Platforms Edition</span>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">-c</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--capture-range</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">none, cudaProfilerApi, hotkey, nvtx</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">When --capture-range is used, profiling will
                                       start only when appropriate start API or hotkey is invoked. If
                                       --capture-range is set to none, start/stop API calls and hotkeys will
                                       be ignored. Note: hotkey works for graphic applications only.
                                       CUDA or NVTX tracing must be enabled on the target application
                                       for '-c cudaProfilerApi' or '-c nvtx' to work.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--export</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">arrow, hdf, json, sqlite, text, none</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Create additional output file(s) based on the data collected.
                                       This option can be given more than once.
                                       WARNING: If the collection captures a large amount of data, creating
                                       the export file may take several minutes to complete.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">-f</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--force-overwrite</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">If true, overwrite all existing result files with same output
                                       filename (.qdstrm, .nsys-rep, .arrows, .hdf, .json, .sqlite, .txt).
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--help</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">&lt;tag&gt;</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Print the help message. The option can take one optional
                                       argument that will be used as a tag. If a tag is provided, only options
                                       relevant to the tag will be printed.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">-o</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--output</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">&lt; filename &gt;</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">report#</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Set report file name. Any %q{ENV_VAR} pattern in the filename
                                       will be substituted with the value of the environment variable. Any %h
                                       pattern in the filename will be substituted with the hostname of the
                                       system. Any %p pattern in the filename will be substituted with the
                                       PID of the target process or the PID of the root process if there is a
                                       process tree. Any %% pattern in the filename will be substituted with %.
                                       Default is report#.{nsys-rep,sqlite,h5,txt,arrows,json} in the working directory.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--session</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">session identifier</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Start the application in the indicated session. The option
                                       argument must represent a valid session name or ID as reported by
                                       <samp class="ph codeph"><span class="ph">nsys</span> sessions list</samp>. Any
                                       <samp class="ph codeph">%q{ENV_VAR}</samp> pattern will be substituted with the
                                       value of the environment variable. Any <samp class="ph codeph">%h</samp> pattern
                                       will be substituted with the hostname of the system. Any <samp class="ph codeph">%%</samp>
                                       pattern will be substituted with <samp class="ph codeph">%</samp>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--session-new</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">[a-Z][0-9,a-Z,spaces]</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">[default]</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Start the application in a new session. Name must start with
                                       an alphabetical character followed by printable or space characters.
                                       Any <samp class="ph codeph">%q{ENV_VAR}</samp> pattern will be substituted with the
                                       value of the environment variable. Any <samp class="ph codeph">%h</samp> pattern
                                       will be substituted with the hostname of the system. Any <samp class="ph codeph">%%</samp>
                                       pattern will be substituted with <samp class="ph codeph">%</samp>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--stats</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Generate summary statistics after the collection. WARNING:
                                       When set to true, an SQLite database will be created after the
                                       collection. If the collection captures a large amount of data,
                                       creating the database file may take several minutes to complete.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">-x</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--stop-on-exit</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">If true, stop collecting automatically when all tracked
                                       processes have exited or when <samp class="ph codeph">stop</samp> command is issued
                                       - whichever occurs first. If false, stop only on <samp class="ph codeph">stop</samp>
                                       command. Note: When this is true, <samp class="ph codeph">stop</samp> command is
                                       optional. <span class="ph">Nsight Systems</span> does not officially support
                                       runs longer than 5 minutes.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--gpu-metrics-device</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">GPU ID, help, all, none</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Collect GPU Metrics from specified devices.
                                       Determine GPU IDs by using <samp class="ph codeph">--gpu-metrics-device=help</samp>
                                       switch.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--gpu-metrics-frequency</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">integer</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">10000</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Specify GPU Metrics sampling frequency. Minimum supported
                                       frequency is 10 (Hz). Maximum supported frequency is 200000(Hz).
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--gpu-metrics-set</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">index</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">first</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Specify metric set for GPU Metrics sampling. The argument
                                       must be one of indices reported by <samp class="ph codeph">--gpu-metrics-set=help</samp>
                                       switch. Default is the first metric set that supports selected GPU.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--gpuctxsw</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">true,false</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Trace GPU context switches. Note that this requires driver
                                       r435.17 or later and root permission. Not supported on IBM Power
                                       targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--accelerator-trace</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">none, nvmedia</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Collect other accelerators workload trace from the hardware
                                       engine units. Available in <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">-b</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--backtrace</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">auto,fp,lbr,dwarf,none</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Select the backtrace method to use while sampling. The option
                                       'lbr' uses Intel(c) Corporation's Last Branch Record registers, available only
                                       with Intel(c) CPUs codenamed Haswell and later. The option 'fp' is frame
                                       pointer and assumes that frame pointers were enabled during compilation.
                                       The option 'dwarf' uses DWARF's CFI (Call Frame Information). Setting the
                                       value to 'none' can reduce collection overhead.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--cpu-core-events (not <span class="ph">Nsight Systems Embedded Platforms Edition</span>)
                                    </td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">'help' or the end users selected events in the format 'x,y'</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">'2' i.e. Instructions Retired</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Select the CPU Core events to sample.  Use the
                                       <samp class="ph codeph">--cpu-core-events=help</samp> switch to see the full list of
                                       events and the number of events that can be collected simultaneously.
                                       Multiple values can be selected, separated by commas only (no spaces).
                                       Use the --event-sample switch to enable.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--cpuctxsw</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">process-tree, system-wide, none</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">process-tree</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Trace OS thread scheduling activity. Select 'none' to disable
                                       tracing CPU context switches. Depending on the platform, some values may
                                       require admin or root privileges. Note: if the --sample switch is set to a value
                                       other than 'none', the --cpuctxsw setting is hardcoded to the same value as the
                                       --sample switch. If --sample=none and a target application is launched, the
                                       default is 'process-tree', otherwise the default is 'none'. Requires
                                       <samp class="ph codeph">--sampling-trigger=perf</samp> switch in <span class="ph">Nsight Systems Embedded Platforms Edition</span>.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--os-events</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">'help' or the end users selected events in the format 'x,y'</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Select the OS events to sample. Use the --os-events=help
                                       switch to see the full list of events. Multiple values can be selected,
                                       separated by commas only (no spaces). Use the --event-sample switch to
                                       enable. Not available in <span class="ph">Nsight Systems Embedded Platforms Edition</span>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--event-sample</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">system-wide, none</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Use the --cpu-core-events=help and the --os-events=help
                                       switches to see the full list of events. If event sampling is enabled
                                       and no events are selected, the CPU Core event 'Instructions Retired' is
                                       selected by default. Not available in <span class="ph">Nsight Systems Embedded Platforms Edition</span>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--event-sampling-frequency</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">Integers from 1 to 20 Hz</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">3</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">The sampling frequency used to collect event counts. Minimum
                                       event sampling frequency is 1 Hz. Maximum event sampling frequency is
                                       20 Hz. Not available in <span class="ph">Nsight Systems Embedded Platforms Edition</span>. 
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--nic-metrics</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Collect metrics from supported NIC/HCA devices</td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">-s</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--sample</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">process-tree, system-wide, none</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">process-tree</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Select how to collect CPU IP/backtrace samples. If 'none' is
                                       selected, CPU sampling is disabled. Depending on the platform, some values may
                                       require admin or root privileges. If a target application is launched, the
                                       default is 'process-tree', otherwise, the default is 'none'. Note: 'system-wide'
                                       is not available on all platforms. Note: If set to 'none', CPU context switch
                                       data will still be collected unless the --cpuctxsw switch is set to 'none'.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--samples-per-backtrace</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">integer &lt;= 32</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">1</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">The number of CPU IP samples collected for every CPU IP/backtrace sample
                                       collected. For example, if set to 4, on the fourth CPU IP sample collected, a
                                       backtrace will also be collected. Lower values increase the amount of data collected.
                                       Higher values can reduce collection overhead and reduce the number of CPU IP
                                       samples dropped. If DWARF backtraces are collected, the default is 4, otherwise the
                                       default is 1. This option is not available on <span class="ph">Nsight Systems Embedded Platforms Edition</span>
                                       or on non-Linux targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--sampling-frequency</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">integers between 100 and 8000</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">1000</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Specify the sampling/backtracing frequency. The minimum
                                       supported frequency is 100 Hz. The maximum supported frequency is 8000
                                       Hz. This option is supported only on QNX, Linux for Tegra, and Windows
                                       targets. Requires <samp class="ph codeph">--sampling-trigger=perf</samp> switch in
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span></td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--sampling-period (<span class="ph">Nsight Systems Embedded Platforms Edition</span>)
                                    </td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">integer</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">determined dynamically</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">The number of CPU Cycle events counted before a CPU
                                       instruction pointer (IP) sample is collected. If configured,
                                       backtraces may also be collected. The smaller the sampling period,
                                       the higher the sampling rate. Note that smaller sampling periods will
                                       increase overhead and significantly increase the size of the result
                                       file(s). Requires <samp class="ph codeph">--sampling-trigger=perf</samp> switch.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--sampling-period (not <span class="ph">Nsight Systems Embedded Platforms Edition</span>)
                                    </td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">integer</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">determined dynamically</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">The number of events counted before a CPU instruction pointer
                                       (IP) sample is collected. The event used to trigger the collection of
                                       a sample is determined dynamically. For example, on Intel based
                                       platforms, it will probably be "Reference Cycles" and on AMD platforms,
                                       "CPU Cycles". If configured, backtraces may also be collected. The
                                       smaller the sampling period, the higher the sampling rate. Note that
                                       smaller sampling periods will increase overhead and significantly
                                       increase the size of the result file(s). This option is available only
                                       on Linux targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--sampling-trigger</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">timer, sched, perf, cuda</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">timer,sched</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Specify backtrace collection trigger. Multiple APIs can be
                                       selected, separated by commas only (no spaces). Available on
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span> targets only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--el1-sampling</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Enable EL1 sampling. Available in
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--el1-sampling-config</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">&lt; filepath config.json &gt;</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">EL1 sampling config. Available in
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span> only. 
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--xhv-trace</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">&lt; filepath pct.json &gt;</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Collect hypervisor trace. Available in
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--xhv-trace-events</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">all, none, core, sched, irq, trap</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">all</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Available in
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--etw-provider</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">"&lt;name&gt;,&lt;guid&gt;", or path to JSON file</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Add custom ETW trace provider(s). If you want to specify more
                                       attributes than Name and GUID, provide a JSON configuration file as
                                       as outlined below. This switch can be used multiple times to add
                                       multiple providers. Note: Only available for Windows targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--retain-etw-files</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Retain ETW files generated by the trace, merge and move the
                                       files to the output directory.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--capture-range-end</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">none, stop, stop-shutdown, repeat[:N], repeat-shutdown:N</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">stop-shutdown</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Specify the desired behavior when a capture range ends.
                                       Applicable only when used along with --capture-range option. If
                                       <samp class="ph codeph">none</samp>, capture range end will be ignored. If
                                       <samp class="ph codeph">stop</samp>, collection will stop at capture range end. Any
                                       subsequent capture ranges will be ignored. Target app will continue
                                       running. If <samp class="ph codeph">stop-shutdown</samp>, collection will stop at
                                       capture range end and session will be shutdown. If
                                       <samp class="ph codeph">repeat[:N]</samp>, collection will stop at capture range end
                                       and subsequent capture ranges will trigger more collections. Use the
                                       optional <samp class="ph codeph">:N</samp> to specify max number of capture ranges
                                       to be honored. Any subsequent capture ranges will be ignored once N
                                       capture ranges are collected. If <samp class="ph codeph">repeat-shutdown:N</samp>,
                                       same behavior as <samp class="ph codeph">repeat:N</samp> but session will be shutdown
                                       after N ranges. For <samp class="ph codeph">stop-shutdown</samp> and
                                       <samp class="ph codeph">repeat-shutdown:N</samp>, as always use --kill option to
                                       specify whether target app should be terminated when shutting down
                                       session.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--ftrace</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Collect ftrace events. Argument should list events to collect
                                       as: subsystem1/event1,subsystem2/event2. Requires root. No ftrace
                                       events are collected by default. Note: Not supported on IBM Power
                                       targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--ftrace-keep-user-config</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Skip initial ftrace setup and collect already configured
                                       events. Default resets the ftrace configuration.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--isr</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Trace Interrupt Service Routines (ISRs) and Deferred
                                       Procedure Calls (DPCs). Requires administrative privileges. Available
                                       only on Windows devices.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--process-scope</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">main, process-tree, system-wide</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">main</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Select which process(es) to trace. Available in
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span> only. <span class="ph">Nsight Systems Workstation Edition</span> will
                                       always trace system-wide in this version of the tool.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--vsync</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Collect vsync events. If collection of vsync events is
                                       enabled, display/display_scanline ftrace events will also be captured.
                                       Available in <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--soc-metrics</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">true,false</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Collect SOC Metrics. Available in
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--soc-metrics-frequency</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">integer</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">100000</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Specify SOC Metrics sampling frequency. Minimum supported
                                       frequency is '100' (Hz). Maximum supported frequency is '1000000'
                                       (Hz). Available in <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e2061" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e2064" class="stentry" rowspan="1" colspan="1">--soc-metrics-set</td>
                                    <td valign="top" headers="d54e2067" class="stentry" rowspan="1" colspan="1">see description</td>
                                    <td valign="top" headers="d54e2070" class="stentry" rowspan="1" colspan="1">see description</td>
                                    <td valign="top" headers="d54e2073" class="stentry" rowspan="1" colspan="1">Specify metric set for SOC Metrics sampling. The option
                                       argument must be one of indices or aliases reported by
                                       <samp class="ph codeph">--soc-metrics-set=help</samp> switch. Default is the first
                                       supported set. Available in <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                              </table>
                           </div>
                        </div>
                     </div>
                     <div class="topic nested2" id="cli-stop-command-switch-options"><a name="cli-stop-command-switch-options" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cli-stop-command-switch-options" name="cli-stop-command-switch-options" shape="rect">1.3.3.&nbsp;CLI Stop Command Switch Options</a></h3>
                        <div class="body">
                           <p class="p">After choosing the <samp class="ph codeph">stop</samp> command switch, the
                              following options are available. Usage:
                           </p><pre xml:space="preserve"><span class="ph">nsys</span> [global-options] stop [options]</pre><div class="p">
                              <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                                 <tr class="sthead">
                                    <th valign="bottom" align="left" id="d54e3001" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Short</th>
                                    <th valign="bottom" align="left" id="d54e3004" class="stentry" width="14.285714285714285%" rowspan="1" colspan="1">Long</th>
                                    <th valign="bottom" align="left" id="d54e3007" class="stentry" width="19.047619047619047%" rowspan="1" colspan="1">Possible Parameters</th>
                                    <th valign="bottom" align="left" id="d54e3010" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Default</th>
                                    <th valign="bottom" align="left" id="d54e3013" class="stentry" width="47.61904761904762%" rowspan="1" colspan="1">Switch Description</th>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3001" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3004" class="stentry" rowspan="1" colspan="1">--help</td>
                                    <td valign="top" headers="d54e3007" class="stentry" rowspan="1" colspan="1">&lt;tag&gt;</td>
                                    <td valign="top" headers="d54e3010" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e3013" class="stentry" rowspan="1" colspan="1">Print the help message. The option can take one optional
                                       argument that will be used as a tag. If a tag is provided, only options
                                       relevant to the tag will be printed.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3001" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3004" class="stentry" rowspan="1" colspan="1">--session</td>
                                    <td valign="top" headers="d54e3007" class="stentry" rowspan="1" colspan="1">session identifier</td>
                                    <td valign="top" headers="d54e3010" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e3013" class="stentry" rowspan="1" colspan="1">Stop the indicated session. The option argument must
                                       represent a valid session name or ID as reported by
                                       <samp class="ph codeph"><span class="ph">nsys</span> sessions list</samp>. Any
                                       <samp class="ph codeph">%q{ENV_VAR}</samp> pattern will be substituted with the
                                       value of the environment variable. Any <samp class="ph codeph">%h</samp> pattern
                                       will be substituted with the hostname of the system. Any <samp class="ph codeph">%%</samp>
                                       pattern will be substituted with <samp class="ph codeph">%</samp>.
                                    </td>
                                 </tr>
                              </table>
                           </div>
                        </div>
                     </div>
                     <div class="topic nested2" id="cli-cancel-command-switch-options"><a name="cli-cancel-command-switch-options" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cli-cancel-command-switch-options" name="cli-cancel-command-switch-options" shape="rect">1.3.4.&nbsp;CLI Cancel Command Switch Options</a></h3>
                        <div class="body">
                           <p class="p">After choosing the <samp class="ph codeph">cancel</samp> command switch, the
                              following options are available. Usage:
                           </p><pre xml:space="preserve"><span class="ph">nsys</span> [global-options] cancel [options]</pre><div class="p">
                              <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                                 <tr class="sthead">
                                    <th valign="bottom" align="left" id="d54e3097" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Short</th>
                                    <th valign="bottom" align="left" id="d54e3100" class="stentry" width="14.285714285714285%" rowspan="1" colspan="1">Long</th>
                                    <th valign="bottom" align="left" id="d54e3103" class="stentry" width="19.047619047619047%" rowspan="1" colspan="1">Possible Parameters</th>
                                    <th valign="bottom" align="left" id="d54e3106" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Default</th>
                                    <th valign="bottom" align="left" id="d54e3109" class="stentry" width="47.61904761904762%" rowspan="1" colspan="1">Switch Description</th>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3097" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3100" class="stentry" rowspan="1" colspan="1">--help</td>
                                    <td valign="top" headers="d54e3103" class="stentry" rowspan="1" colspan="1">&lt;tag&gt;</td>
                                    <td valign="top" headers="d54e3106" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e3109" class="stentry" rowspan="1" colspan="1">Print the help message. The option can take one optional
                                       argument that will be used as a tag. If a tag is provided, only options
                                       relevant to the tag will be printed.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3097" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3100" class="stentry" rowspan="1" colspan="1">--session</td>
                                    <td valign="top" headers="d54e3103" class="stentry" rowspan="1" colspan="1">&lt;session identifier&gt;</td>
                                    <td valign="top" headers="d54e3106" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e3109" class="stentry" rowspan="1" colspan="1">Cancel the collection in the given session. The option argument must represent a
                                       valid session name or ID as reported by
                                       <samp class="ph codeph"><span class="ph">nsys</span> sessions list</samp>. Any <samp class="ph codeph">%q{ENV_VAR}</samp>
                                       pattern in the option argument will be substituted with the value of the environment
                                       variable. Any <samp class="ph codeph">%h</samp> pattern in the option argument will be substituted with
                                       the hostname of the system. Any <samp class="ph codeph">%%</samp> pattern in the option argument will
                                       be substituted with <samp class="ph codeph">%</samp>.
                                    </td>
                                 </tr>
                              </table>
                           </div>
                        </div>
                     </div>
                     <div class="topic nested2" id="cli-launch-command-switch-options"><a name="cli-launch-command-switch-options" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cli-launch-command-switch-options" name="cli-launch-command-switch-options" shape="rect">1.3.5.&nbsp;CLI Launch Command Switch Options</a></h3>
                        <div class="body">
                           <p class="p">After choosing the <samp class="ph codeph">launch</samp> command switch, the following options are
                              available. Usage:
                           </p><pre xml:space="preserve"><span class="ph">nsys</span> [global-options] launch [options] &lt;application&gt; [application-arguments]</pre><div class="p">
                              <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                                 <tr class="sthead">
                                    <th valign="bottom" align="left" id="d54e3192" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Short</th>
                                    <th valign="bottom" align="left" id="d54e3195" class="stentry" width="14.285714285714285%" rowspan="1" colspan="1">Long</th>
                                    <th valign="bottom" align="left" id="d54e3198" class="stentry" width="19.047619047619047%" rowspan="1" colspan="1">Possible Parameters</th>
                                    <th valign="bottom" align="left" id="d54e3201" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Default</th>
                                    <th valign="bottom" align="left" id="d54e3204" class="stentry" width="47.61904761904762%" rowspan="1" colspan="1">Switch Description</th>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">-e</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--env-var</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">A=B</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">NA</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Set environment variable(s) for the application process to be
                                       launched. Environment variables should be defined as A=B. Multiple
                                       environment variables can be specified as A=B,C=D.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--help</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">&lt;tag&gt;</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Print the help message. The option can take one optional
                                       argument that will be used as a tag. If a tag is provided, only options
                                       relevant to the tag will be printed.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">-n</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--inherit-environment</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">When true, the current environment variables and the tool’s
                                       environment variables will be specified for the launched process. When
                                       false, only the tool’s environment variables will be specified for the
                                       launched process.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--session</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">session identifier</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Launch the application in the indicated session. The option
                                       argument must represent a valid session name or ID as reported by
                                       <samp class="ph codeph"><span class="ph">nsys</span> sessions list</samp>. Any
                                       <samp class="ph codeph">%q{ENV_VAR}</samp> pattern will be substituted with the
                                       value of the environment variable. Any <samp class="ph codeph">%h</samp> pattern
                                       will be substituted with the hostname of the system. Any <samp class="ph codeph">%%</samp>
                                       pattern will be substituted with <samp class="ph codeph">%</samp>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--session-new</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">[a-Z][0-9,a-Z,spaces]</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">[default]</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Launch the application in a new session. Name must start with
                                       an alphabetical character followed by printable or space characters.
                                       Any <samp class="ph codeph">%q{ENV_VAR}</samp> pattern will be substituted with the
                                       value of the environment variable. Any <samp class="ph codeph">%h</samp> pattern will
                                       be substituted with the hostname of the system. Any <samp class="ph codeph">%%</samp>
                                       pattern will be substituted with <samp class="ph codeph">%</samp>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">-w</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--show-output</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">If true, send target process's stdout and stderr streams to both the console and
                                       stdout/stderr files which are added to the QDSTRM file.
                                       If false, only send target process stdout and stderr streams to the stdout/stderr files
                                       which are added to the QDSTRM file.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">-t</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--trace</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">cuda, nvtx, cublas, cublas-verbose, cusparse, cusparse-verbose,
                                       cudnn, opengl, opengl-annotations, openacc, openmp, osrt,
                                       mpi, nvvideo, vulkan, vulkan-annotations, dx11, dx11-annotations, dx12,
                                       dx12-annotations, oshmem, ucx, wddm, nvmedia, none
                                    </td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">cuda, opengl, nvtx, osrt</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Select the API(s) to be traced. The osrt switch controls the
                                       OS runtime libraries tracing. Multiple APIs can be selected, separated
                                       by commas only (no spaces). Since OpenACC, cuDNN and cuBLAS APIs are
                                       tightly linked with CUDA, selecting one of those APIs will automatically
                                       enable CUDA tracing. Reflex SDK latency markers will be automatically
                                       collected when DX or vulkan API trace is enabled. See information on
                                       --mpi-impl option below if mpi is selected. If '&lt;api&gt;-annotations'
                                       is selected, the corresponding API will also be traced. If the none
                                       option is selected, no APIs are traced and no other API can be selected.
                                       Note: cublas, cudnn, nvvideo, opengl, and vulkan are
                                       not available on IBM Power target.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--dx-force-declare-adapter-removal-support</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">The <span class="ph">Nsight Systems</span> trace initialization involves creating a D3D
                                       device and discarding it. Enabling this flag makes a call to
                                       DXGIDeclareAdapterRemovalSupport() before device creation.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--dx12-gpu-workload</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true, false, individual, batch, none</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">individual</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">If individual or true, trace each DX12 workload's GPU activity individually.
                                       If batch, trace DX12 workloads' GPU activity in ExecuteCommandLists call batches.
                                       If none or false, do not trace DX12 workloads' GPU activity.
                                       Note that this switch is applicable only when --trace=dx12 is specified.
                                       This option is only supported on Windows targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--dx12-wait-calls</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">If true, trace wait calls that block on fences for DX12. Note
                                       that this switch is applicable only when --trace=dx12 is specified. This
                                       option is only supported on Windows targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--opengl-gpu-workload</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">If true, trace the OpenGL workloads' GPU activity. Note that
                                       this switch is applicable only when --trace=opengl is specified. This
                                       option is not supported on IBM Power targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--vulkan-gpu-workload</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true, false, individual, batch, none</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">individual</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">If individual or true, trace each Vulkan workload's GPU activity individually.
                                       If batch, trace Vulkan workloads' GPU activity in vkQueueSubmit call batches.
                                       If none or false, do not trace Vulkan workloads' GPU activity.
                                       Note that this switch is applicable only when --trace=vulkan is specified. This option
                                       is not supported on QNX.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--wddm-additional-events</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">If true, collect additional range of ETW events, including
                                       context status, allocations, sync wait and signal events, etc. Note
                                       that this switch is applicable only when --trace=wddm is specified.
                                       This option is only supported on Windows targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">-b</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--backtrace</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">WARNING: This switch is no longer supported.
                                       Please set the --backtrace switch when using the start command instead.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--clock-frequency-changes</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Collect clock frequency changes. Available in
                                       <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--cpu-cluster-events</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">0x16, 0x17, ..., none</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Collect per-cluster Uncore PMU counters. Multiple values can
                                       be selected, separated by commas only (no spaces). Use the
                                       --cpu-cluster-events=help switch to see the full list of values.
                                       Available in <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--cpu-core-events (<span class="ph">Nsight Systems Embedded Platforms Edition</span>)
                                    </td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">0x11,0x13,...,none</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Collect per-core PMU counters. Multiple values can be
                                       selected, separated by commas only (no spaces). Use the
                                       <samp class="ph codeph">--cpu-core-events=help</samp> switch to see the full list of
                                       values.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--cpu-socket-events</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">0x2a, 0x2c, ..., none</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Collect per-socket Uncore PMU counters. Multiple values can
                                       be selected, separated by commas only (no spaces). Use the
                                       --cpu-socket-events=help switch to see the full list of values.
                                       Available in <span class="ph">Nsight Systems Embedded Platforms Edition</span> only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--cpuctxsw</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">WARNING: This switch is no longer supported.
                                       Please set the --cpuctxsw switch when using the start command instead.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--nic-metrics</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Collect metrics from supported NIC/HCA devices</td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--osrt-backtrace-depth</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">integer</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">24</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Set the depth for the backtraces collected for OS runtime
                                       libraries calls.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--osrt-backtrace-stack-size</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">integer</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">6144</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Set the stack dump size, in bytes, to generate
                                       backtraces for OS runtime libraries calls.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--osrt-backtrace-threshold</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">nanoseconds</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">80000</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Set the duration, in nanoseconds, that all OS runtime
                                       libraries calls must execute before backtraces are collected. 
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--osrt-threshold</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">&lt; nanoseconds &gt;</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">1000 ns</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Set the duration, in nanoseconds, that Operating System Runtime (osrt) APIs must
                                       execute before they are traced. Values much less than 1000 may cause significant overhead
                                       and result in extremely large result files. Default is 1000 (1 microsecond). Note:
                                       Not available for IBM Power targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--qnx-kernel-events</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">class/event,event,class/event:mode,class:mode,help,none</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Multiple values can be selected, separated by commas only (no spaces). See
                                       the --qnx-kernel-events-mode switch description for ':mode' format. Use the
                                       '--qnx-kernel-events=help' switch to see the full list of values.
                                       Example: '--qnx-kernel-events=8/1:system:wide,_NTO_TRACE_THREAD:process:fast,_NTO_TRACE_KERCALLENTER/__KER_BAD,_NTO_TRACE_COMM,13'.
                                       Collect QNX kernel events.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--qnx-kernel-events-mode</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">system,process,fast,wide</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">system:fast</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Values are separated by a colon (':') only (no spaces). 'system' and 'process'
                                       cannot be specified at the same time. 'fast' and 'wide' cannot be specified at the same
                                       time. Please check the QNX documentation to determine when to select the 'fast' or 'wide'
                                       mode. Specify the default mode for QNX kernel events collection.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--resolve-symbols</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true,false</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Resolve symbols of captured samples and backtraces.</td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--injection-use-detours</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true,false</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Use detours for injection. If false, process injection will be performed by 
                                       windows hooks which allows to bypass anti-cheat software.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--samples-per-backtrace</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">WARNING: This switch is no longer supported.
                                       Please set the --samples-per-backtrace switch when using the start command
                                       instead.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--sampling-frequency</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">WARNING: This switch is no longer supported.
                                       Please set the --sampling-frequency switch when using the start command
                                       instead.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--sampling-period</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">WARNING: This switch is no longer supported.
                                       Please set the --sampling-period switch when using the start command instead.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--sampling-trigger</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">WARNING: This switch is no longer supported.
                                       Please set the --sampling-trigger switch when using the start command instead.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--cuda-flush-interval</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">milliseconds</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">See description</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Set the interval, in milliseconds, when buffered CUDA data is automatically
                                       saved to storage. CUDA data buffer saves may cause profiler overhead. Buffer save
                                       behavior can be controlled with this switch.
                                       
                                       If the CUDA flush interval is set to 0 on systems running CUDA 11.0 or newer, buffers are
                                       saved when they fill. If a flush interval is set to a non-zero value on such systems,
                                       buffers are saved only when the flush interval expires. If a flush interval is set and
                                       the profiler runs out of available buffers before the flush interval expires, additional
                                       buffers will be allocated as needed. In this case, setting a flush interval can reduce
                                       buffer save overhead but increase memory use by the profiler.
                                       
                                       If the flush interval is set to 0 on systems running older versions of CUDA,
                                       buffers are saved at the end of the collection. If the profiler runs out of available
                                       buffers, additional buffers are allocated as needed. If a flush interval is set to a
                                       non-zero value on such systems, buffers are saved when the flush interval expires.
                                       A cuCtxSynchronize call may be inserted into the workflow before the buffers
                                       are saved which will cause application overhead. In this case, setting a flush interval
                                       can reduce memory use by the profiler but may increase save overhead.
                                       
                                       For collections over 30 seconds an interval of 10 seconds is recommended.
                                       Default is 10000 for <span class="ph">Nsight Systems Embedded Platforms Edition</span> and 0 otherwise.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--cuda-memory-usage</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Track the GPU memory usage by CUDA kernels. Applicable only
                                       when CUDA tracing is enabled. Note: This feature may cause significant
                                       runtime overhead.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--cuda-um-cpu-page-faults</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">This switch tracks the page faults that occur
                                       when CPU code tries to access a memory page that resides on the device.
                                       Note that this feature may cause significant runtime overhead. Not
                                       available on <span class="ph">Nsight Systems Embedded Platforms Edition</span>.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--cuda-um-gpu-page-faults</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">This switch tracks the page faults that occur
                                       when GPU code tries to access a memory page that resides on the host.
                                       Note that this feature may cause significant runtime overhead. Not
                                       available on <span class="ph">Nsight Systems Embedded Platforms Edition</span>. 
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--cudabacktrace</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">all, none, kernel, memory, sync, other</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">When tracing CUDA APIs, enable the collection of a backtrace
                                       when a CUDA API is invoked. Significant runtime overhead may occur.
                                       Values may be combined using ','. Each value except 'none' may be
                                       appended with a threshold after ':'. Threshold is duration, in
                                       nanoseconds, that CUDA APIs must execute before backtraces are
                                       collected, e.g. 'kernel:500'. Default value for each threshold is
                                       1000ns (1us). Note: CPU sampling must be enabled. Note: Not
                                       available on IBM Power targets.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--cuda-graph-trace</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">graph, node</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">graph</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">If 'graph' is selected, CUDA graphs will be traced as a whole
                                       and node activities will not be collected. This will reduce overhead
                                       to a minimum, but requires CUDA driver version 515.43 or higher.
                                       If 'node' is selected, node activities will be collected, but CUDA graphs
                                       will not be traced as a whole. This may cause significant runtime overhead.
                                       Default is 'graph' if available, otherwise default is 'node'.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--nvtx-domain-include</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">default, &lt;domain_names&gt;</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Choose to only include NVTX events from a comma separated
                                       list of domains. 'default' filters the NVTX default domain. A domain
                                       with this name or commas in a domain name must be escaped with '\'.
                                       Note: Only one of --nvtx-domain-include and --nvtx-domain-exclude can
                                       be used. This option is only applicable when --trace=nvtx is specified.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--nvtx-domain-exclude</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">default, &lt;domain_names&gt;</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Choose to exclude NVTX events from a comma separated
                                       list of domains. 'default' filters the NVTX default domain. A domain
                                       with this name or commas in a domain name must be escaped with '\'.
                                       Note: Only one of --nvtx-domain-include and --nvtx-domain-exclude can
                                       be used. This option is only applicable when --trace=nvtx is specified.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">-p</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--nvtx-capture</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">range@domain, range, range@*</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Specify NVTX range and domain to trigger the profiling session.
                                       Note that this switch is applicable only when --capture-range=nvtx is specified at the
                                       start of the profiled session.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--command-file</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">&lt; filename &gt;</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Open a file that contains launch switches and parse the
                                       switches. Note additional switches on the command line will override
                                       switches in the file. This flag can be specified more than once.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--hotkey-capture</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">'F1' to 'F12'</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">'F12'</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Hotkey to trigger the profiling session. Note that this switch is applicable
                                       only when --capture-range=hotkey is specified at the start of the profiled
                                       session.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--isr</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true,false</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">Trace Interrupt Service Routines (ISRs) and Deferred Procedure Calls (DPCs).
                                       Requires administrative privileges. Available only on Windows devices.
                                    </td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">false</td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--mpi-impl</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">openmpi,mpich</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">openmpi</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">When using --trace=mpi to trace MPI APIs use --mpi-impl to
                                       specify which MPI implementation the application is using. If no MPI implementation is
                                       specified, <span class="ph">nsys</span> tries to automatically detect it based on the dynamic
                                       linker's search path. If this fails, 'openmpi' is used. Calling --mpi-impl without
                                       --trace=mpi is not supported.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--run-as</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">&lt; username &gt;</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">Run the target application as the specified username. If not
                                       specified, the target application will be run by the same user as
                                       <span class="ph">Nsight Systems</span>. Requires root privileges. Available for
                                       Linux targets only. 
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">-s</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--sample</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">WARNING: This switch is no longer supported.
                                       Please set the --sample switch when using the start command instead.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--trace-fork-before-exec</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">If true, trace any child process after fork and before they
                                       call one of the exec functions. Beware, tracing in this interval
                                       relies on undefined behavior and might cause your application to
                                       crash or deadlock. Note: This option is only available on Linux target
                                       platforms.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e3192" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e3195" class="stentry" rowspan="1" colspan="1">--wait</td>
                                    <td valign="top" headers="d54e3198" class="stentry" rowspan="1" colspan="1">primary,all</td>
                                    <td valign="top" headers="d54e3201" class="stentry" rowspan="1" colspan="1">all</td>
                                    <td valign="top" headers="d54e3204" class="stentry" rowspan="1" colspan="1">If primary, the CLI will wait on the application process
                                       termination. If all, the CLI will additionally wait on re-parented
                                       processes created by the application.
                                    </td>
                                 </tr>
                              </table>
                           </div>
                        </div>
                     </div>
                     <div class="topic nested2" id="cli-shutdown-command-switch-options"><a name="cli-shutdown-command-switch-options" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cli-shutdown-command-switch-options" name="cli-shutdown-command-switch-options" shape="rect">1.3.6.&nbsp;CLI Shutdown Command Switch Options</a></h3>
                        <div class="body">
                           <p class="p">After choosing the <samp class="ph codeph">shutdown</samp> command switch, the
                              following options are available. Usage:
                           </p><pre xml:space="preserve"><span class="ph">nsys</span> [global-options] shutdown [options]</pre><div class="p">
                              <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                                 <tr class="sthead">
                                    <th valign="bottom" align="left" id="d54e4130" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Short</th>
                                    <th valign="bottom" align="left" id="d54e4133" class="stentry" width="14.285714285714285%" rowspan="1" colspan="1">Long</th>
                                    <th valign="bottom" align="left" id="d54e4136" class="stentry" width="19.047619047619047%" rowspan="1" colspan="1">Possible Parameters</th>
                                    <th valign="bottom" align="left" id="d54e4139" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Default</th>
                                    <th valign="bottom" align="left" id="d54e4142" class="stentry" width="47.61904761904762%" rowspan="1" colspan="1">Switch Description</th>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4130" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4133" class="stentry" rowspan="1" colspan="1">--help</td>
                                    <td valign="top" headers="d54e4136" class="stentry" rowspan="1" colspan="1">&lt;tag&gt;</td>
                                    <td valign="top" headers="d54e4139" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e4142" class="stentry" rowspan="1" colspan="1">Print the help message. The option can take one optional
                                       argument that will be used as a tag. If a tag is provided, only options
                                       relevant to the tag will be printed.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4130" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4133" class="stentry" rowspan="1" colspan="1">--kill</td>
                                    <td valign="top" headers="d54e4136" class="stentry" rowspan="1" colspan="1">
                                       <p class="p">On Linux:
                                          one, sigkill, sigterm, signal number
                                       </p>
                                       <p class="p">On Windows:
                                          true, false
                                       </p>
                                    </td>
                                    <td valign="top" headers="d54e4139" class="stentry" rowspan="1" colspan="1">
                                       <p class="p">On Linux: sigterm</p>
                                       <p class="p">On Windows: true</p>
                                    </td>
                                    <td valign="top" headers="d54e4142" class="stentry" rowspan="1" colspan="1">Send signal to the target application's process group when shutting down
                                       session.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4130" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4133" class="stentry" rowspan="1" colspan="1">--session</td>
                                    <td valign="top" headers="d54e4136" class="stentry" rowspan="1" colspan="1">session identifier</td>
                                    <td valign="top" headers="d54e4139" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e4142" class="stentry" rowspan="1" colspan="1">Shutdown the indicated session. The option argument must
                                       represent a valid session name or ID as reported by
                                       <samp class="ph codeph"><span class="ph">nsys</span> sessions list</samp>. Any
                                       <samp class="ph codeph">%q{ENV_VAR}</samp> pattern will be substituted with the
                                       value of the environment variable. Any <samp class="ph codeph">%h</samp> pattern
                                       will be substituted with the hostname of the system. Any <samp class="ph codeph">%%</samp>
                                       pattern will be substituted with <samp class="ph codeph">%</samp>.
                                    </td>
                                 </tr>
                              </table>
                           </div>
                        </div>
                     </div>
                     <div class="topic nested2" id="cli-export-command-switch-options"><a name="cli-export-command-switch-options" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cli-export-command-switch-options" name="cli-export-command-switch-options" shape="rect">1.3.7.&nbsp;CLI Export Command Switch Options</a></h3>
                        <div class="body">
                           <p class="p">After choosing the <samp class="ph codeph">export</samp> command switch, the
                              following options are available. Usage:
                           </p><pre xml:space="preserve"><span class="ph">nsys</span> [global-options] export [options] [nsys-rep-file]</pre><div class="p">
                              <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                                 <tr class="sthead">
                                    <th valign="bottom" align="left" id="d54e4254" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Short</th>
                                    <th valign="bottom" align="left" id="d54e4257" class="stentry" width="14.285714285714285%" rowspan="1" colspan="1">Long</th>
                                    <th valign="bottom" align="left" id="d54e4260" class="stentry" width="19.047619047619047%" rowspan="1" colspan="1">Possible Parameters</th>
                                    <th valign="bottom" align="left" id="d54e4263" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Default</th>
                                    <th valign="bottom" align="left" id="d54e4266" class="stentry" width="47.61904761904762%" rowspan="1" colspan="1">Switch Description</th>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4254" class="stentry" rowspan="1" colspan="1">-f</td>
                                    <td valign="top" headers="d54e4257" class="stentry" rowspan="1" colspan="1">--force-overwrite</td>
                                    <td valign="top" headers="d54e4260" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e4263" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e4266" class="stentry" rowspan="1" colspan="1">If true, overwrite all existing result files with same output
                                       filename (QDSTRM, nsys-rep, SQLITE, HDF, TEXT, ARROW, JSON).
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4254" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4257" class="stentry" rowspan="1" colspan="1">--help</td>
                                    <td valign="top" headers="d54e4260" class="stentry" rowspan="1" colspan="1">&lt;tag&gt;</td>
                                    <td valign="top" headers="d54e4263" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e4266" class="stentry" rowspan="1" colspan="1">Print the help message. The option can take one optional
                                       argument that will be used as a tag. If a tag is provided, only options
                                       relevant to the tag will be printed.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4254" class="stentry" rowspan="1" colspan="1">-o</td>
                                    <td valign="top" headers="d54e4257" class="stentry" rowspan="1" colspan="1">--output</td>
                                    <td valign="top" headers="d54e4260" class="stentry" rowspan="1" colspan="1">&lt;filename&gt;</td>
                                    <td valign="top" headers="d54e4263" class="stentry" rowspan="1" colspan="1">&lt;inputfile.ext&gt;</td>
                                    <td valign="top" headers="d54e4266" class="stentry" rowspan="1" colspan="1">Set the .output filename. The default is the input filename
                                       with the extension for the chosen format.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4254" class="stentry" rowspan="1" colspan="1">-t</td>
                                    <td valign="top" headers="d54e4257" class="stentry" rowspan="1" colspan="1">--type</td>
                                    <td valign="top" headers="d54e4260" class="stentry" rowspan="1" colspan="1">arrow, hdf, info, json, sqlite, text</td>
                                    <td valign="top" headers="d54e4263" class="stentry" rowspan="1" colspan="1">sqlite</td>
                                    <td valign="top" headers="d54e4266" class="stentry" rowspan="1" colspan="1">Export format type. HDF format is supported only on x86_64
                                       Linux and Windows
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4254" class="stentry" rowspan="1" colspan="1">-l</td>
                                    <td valign="top" headers="d54e4257" class="stentry" rowspan="1" colspan="1">--lazy</td>
                                    <td valign="top" headers="d54e4260" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e4263" class="stentry" rowspan="1" colspan="1">true</td>
                                    <td valign="top" headers="d54e4266" class="stentry" rowspan="1" colspan="1"> Controls if table creation is lazy or not. When true, a
                                       table will only be created when it contains data. This option
                                       will be deprecated in the future, and all exports will be non-lazy.
                                       This affects SQLite, HDF5, and Arrow exports only.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4254" class="stentry" rowspan="1" colspan="1">-q</td>
                                    <td valign="top" headers="d54e4257" class="stentry" rowspan="1" colspan="1">--quiet</td>
                                    <td valign="top" headers="d54e4260" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e4263" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e4266" class="stentry" rowspan="1" colspan="1">If true, do not display progress bar</td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4254" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4257" class="stentry" rowspan="1" colspan="1">--separate-strings</td>
                                    <td valign="top" headers="d54e4260" class="stentry" rowspan="1" colspan="1">true,false</td>
                                    <td valign="top" headers="d54e4263" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e4266" class="stentry" rowspan="1" colspan="1">Output stored strings and thread names separately, with one
                                       value per line. This affects JSON and text output only.
                                    </td>
                                 </tr>
                              </table>
                           </div>
                        </div>
                     </div>
                     <div class="topic nested2" id="cli-stats-command-switch-options"><a name="cli-stats-command-switch-options" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cli-stats-command-switch-options" name="cli-stats-command-switch-options" shape="rect">1.3.8.&nbsp;CLI Stats Command Switch Options</a></h3>
                        <div class="body">
                           <p class="p">
                              The <samp class="ph codeph"><span class="ph">nsys</span> stats</samp> command generates
                              a series of summary or trace reports. These reports can be output to the
                              console, or to individual files, or piped to external processes. Reports
                              can be rendered in a variety of different output formats, from human
                              readable columns of text, to formats more appropriate for data exchange,
                              such as CSV.
                           </p>
                           <p class="p">
                              Reports are generated from an SQLite export of a .nsys-rep file. If a .nsys-rep
                              file is specified, <span class="ph">Nsight Systems</span> will look for an
                              accompanying SQLite file and use it. If no SQLite file exists, one will
                              be exported and created.
                           </p>
                           <p class="p">
                              Individual reports are generated by calling out to scripts that read data
                              from the SQLite file and return their report data in CSV format.
                              <span class="ph">Nsight Systems</span> ingests this data and formats it as
                              requested, then displays the data to the console, writes it to a file, or
                              pipes it to an external process. Adding new reports is as simple as
                              writing a script that can read the SQLite file and generate the required
                              CSV output. See the shipped scripts as an example. Both reports and
                              formatters may take arguments to tweak their processing. For details on
                              shipped scripts and formatters, see <strong class="ph b">Report Scripts</strong> topic.
                           </p>
                           <p class="p">
                              Reports are processed using a three-tuple that consists of 1) the requested
                              report (and any arguments), 2) the presentation format (and any arguments),
                              and 3) the output (filename, console, or external process). The first
                              report specified uses the first format specified, and is presented via
                              the first output specified. The second report uses the second format for
                              the second output, and so forth. If more reports are specified than
                              formats or outputs, the format and/or output list is expanded to match
                              the number of provided reports by repeating the last specified element of
                              the list (or the default, if nothing was specified).
                           </p>
                           <p class="p"><samp class="ph codeph"><span class="ph">nsys</span> stats</samp> is a very powerful
                              command and can handle complex argument structures, please see the topic
                              below on Example Stats Command Sequences.
                           </p>
                           <p class="p">After choosing the <samp class="ph codeph">stats</samp> command switch, the
                              following options are available. Usage:
                           </p>
                           <p class="p"><samp class="ph codeph"><span class="ph">nsys</span> [global-options] stats [options] [input-file]
                                 </samp></p>
                           <div class="p">
                              <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                                 <tr class="sthead">
                                    <th valign="bottom" align="left" id="d54e4460" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Short</th>
                                    <th valign="bottom" align="left" id="d54e4463" class="stentry" width="14.285714285714285%" rowspan="1" colspan="1">Long</th>
                                    <th valign="bottom" align="left" id="d54e4466" class="stentry" width="19.047619047619047%" rowspan="1" colspan="1">Possible Parameters</th>
                                    <th valign="bottom" align="left" id="d54e4469" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Default</th>
                                    <th valign="bottom" align="left" id="d54e4472" class="stentry" width="47.61904761904762%" rowspan="1" colspan="1">Switch Description</th>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4460" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4463" class="stentry" rowspan="1" colspan="1">--help</td>
                                    <td valign="top" headers="d54e4466" class="stentry" rowspan="1" colspan="1">&lt;tag&gt;</td>
                                    <td valign="top" headers="d54e4469" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e4472" class="stentry" rowspan="1" colspan="1">Print the help message. The option can take one optional
                                       argument that will be used as a tag. If a tag is provided, only options
                                       relevant to the tag will be printed.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4460" class="stentry" rowspan="1" colspan="1">-f</td>
                                    <td valign="top" headers="d54e4463" class="stentry" rowspan="1" colspan="1">--format</td>
                                    <td valign="top" headers="d54e4466" class="stentry" rowspan="1" colspan="1">column, table, csv, tsv, json, hdoc, htable, .</td>
                                    <td valign="top" headers="d54e4469" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4472" class="stentry" rowspan="1" colspan="1"> Specify the output format of the corresponding report(s). The
                                       special name "." indicates the default format for the given
                                       output. The default format for console is column, while files and
                                       process outputs default to csv. This option may be used multiple times.
                                       Multiple formats may also be specified using a comma-separated list
                                       (&lt;name[:args...][,name[:args...]...]&gt;). See
                                       <strong class="ph b">Report Scripts</strong> for options
                                       available with each format.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4460" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4463" class="stentry" rowspan="1" colspan="1">--force-export</td>
                                    <td valign="top" headers="d54e4466" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e4469" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e4472" class="stentry" rowspan="1" colspan="1">Force a re-export of the SQLite file from the specified
                                       .nsys-rep file, even if an SQLite file already exists.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4460" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4463" class="stentry" rowspan="1" colspan="1">--force-overwrite</td>
                                    <td valign="top" headers="d54e4466" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e4469" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e4472" class="stentry" rowspan="1" colspan="1">Overwrite any existing report file(s).</td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4460" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4463" class="stentry" rowspan="1" colspan="1">--help-formats</td>
                                    <td valign="top" headers="d54e4466" class="stentry" rowspan="1" colspan="1">&lt;format_name&gt;, ALL, [none]</td>
                                    <td valign="top" headers="d54e4469" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e4472" class="stentry" rowspan="1" colspan="1">With no argument, give a summary of the available output
                                       formats.  If a format name is given, a more detailed explanation of
                                       that format is displayed.  If <samp class="ph codeph">ALL</samp> is
                                       given, a more detailed explanation of all available formats is
                                       displayed.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4460" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4463" class="stentry" rowspan="1" colspan="1">--help-reports</td>
                                    <td valign="top" headers="d54e4466" class="stentry" rowspan="1" colspan="1">&lt;report_name&gt;, ALL, [none]</td>
                                    <td valign="top" headers="d54e4469" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e4472" class="stentry" rowspan="1" colspan="1">With no argument, list a summary of the available summary and
                                       trace reports.  If a report name is given, a more detailed
                                       explanation of the report is displayed.  If <samp class="ph codeph">ALL</samp> is
                                       given, a more detailed explanation of all available reports is
                                       displayed.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4460" class="stentry" rowspan="1" colspan="1">-o</td>
                                    <td valign="top" headers="d54e4463" class="stentry" rowspan="1" colspan="1">--output</td>
                                    <td valign="top" headers="d54e4466" class="stentry" rowspan="1" colspan="1">-, @&lt;command&gt;, &lt;basename&gt;, .</td>
                                    <td valign="top" headers="d54e4469" class="stentry" rowspan="1" colspan="1">-</td>
                                    <td valign="top" headers="d54e4472" class="stentry" rowspan="1" colspan="1">Specify the output mechanism for the corresponding rule(s).
                                       There are three output mechanisms: print to console, output to
                                       file, or output to command.
                                       
                                       This option may be used multiple times. Multiple outputs may also
                                       be specified using a comma-separated list.
                                       
                                       If the given output name is "-", the rule will be displayed
                                       on the console. If the output name starts with "@", the output
                                       designates a command to run. The command will be executed and
                                       the rule output will be piped into the command. Any other output
                                       is assumed to be the base path and name for a file.
                                       
                                       If a file basename is given, the filename used will be:
                                       
                                       &lt;basename&gt;_&lt;report&amp;args&gt;.&lt;output_format&gt;.
                                       
                                       The default base (including path) is the name of the SQLite file
                                       (as derived from the input file or --sqlite option), minus the
                                       extension. The output "." can be used to indicate the rule
                                       should be output to a file, and the default basename should be
                                       used. To write one or more rules to files using the default
                                       basename, use the option: "--output .".
                                       
                                       If the output starts with "@", the rule is output to the
                                       given command. The command is run, and the output of the rule
                                       is piped to the command's stdin (standard-input). The command's
                                       stdout and stderr remain attached to the console, so any output
                                       will be displayed directly to the console.
                                       
                                       Be aware there are some limitations in how the command string is
                                       parsed. No shell expansions (including *, ?, [], and ~) are
                                       supported. The command cannot be piped to another command, nor
                                       redirected to a file using shell syntax. The command and command
                                       arguments are split on whitespace, and no quotes (within the command
                                       syntax) are supported. For commands that require complex command
                                       line syntax, it is suggested that the command be put into a shell
                                       script file, and the script designated as the output command.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4460" class="stentry" rowspan="1" colspan="1">-q</td>
                                    <td valign="top" headers="d54e4463" class="stentry" rowspan="1" colspan="1">--quiet</td>
                                    <td valign="top" headers="d54e4466" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4469" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4472" class="stentry" rowspan="1" colspan="1">Do not display verbose messages, only display errors.</td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4460" class="stentry" rowspan="1" colspan="1">-r</td>
                                    <td valign="top" headers="d54e4463" class="stentry" rowspan="1" colspan="1">--report</td>
                                    <td valign="top" headers="d54e4466" class="stentry" rowspan="1" colspan="1">See <strong class="ph b">Report Scripts</strong></td>
                                    <td valign="top" headers="d54e4469" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4472" class="stentry" rowspan="1" colspan="1">Specify the report(s) to generate, including any arguments.
                                       This option may be used multiple times. Multiple reports may also be
                                       specified using a comma-separated list
                                       (&lt;name[:args...][,name[:args...]...]&gt;). If no reports are
                                       specified, the following will be used as the default report set:
                                       nvtxsum, osrtsum, cudaapisum, gpukernsum, gpumemtimesum, gpumemsizesum,
                                       openmpevtsum, khrdebugsum, khrdebuggpusum, vulkanmarkerssum,
                                       vulkangpumarkersum, dx11pixsum, dx12gpumarkersum, dx12pixsum,
                                       wddmqueuesdetails, unifiedmemory, unifiedmemorytotals, umcpupagefaults,
                                       openaccsum. See <strong class="ph b">Report Scripts</strong> section for details about
                                       existing built-in scripts and how to make your own.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4460" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4463" class="stentry" rowspan="1" colspan="1">--report-dir</td>
                                    <td valign="top" headers="d54e4466" class="stentry" rowspan="1" colspan="1">&lt;path&gt;</td>
                                    <td valign="top" headers="d54e4469" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4472" class="stentry" rowspan="1" colspan="1">Add a directory to the path used to find report scripts. This
                                       is usually only needed if you have one or more directories with
                                       personal scripts. This option may be used multiple times. Each use
                                       adds a new directory to the end of the path. A search path can also
                                       be defined using the environment variable "NSYS_STATS_REPORT_PATH".
                                       Directories added this way will be added after the application flags.
                                       The last two entries in the path will always be the current working
                                       directory, followed by the directory containing the shipped
                                       <samp class="ph codeph"><span class="ph">nsys</span></samp> reports.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4460" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4463" class="stentry" rowspan="1" colspan="1">--sqlite</td>
                                    <td valign="top" headers="d54e4466" class="stentry" rowspan="1" colspan="1">&lt;file.sqlite&gt;</td>
                                    <td valign="top" headers="d54e4469" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4472" class="stentry" rowspan="1" colspan="1">Specify the SQLite export filename. If this file exists, it
                                       will be used. If this file doesn't exist (or if --force-export was
                                       given) this file will be created from the specified .nsys-rep file
                                       before report processing. This option cannot be used if the specified
                                       input file is also an SQLite file.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4460" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4463" class="stentry" rowspan="1" colspan="1">--timeunit</td>
                                    <td valign="top" headers="d54e4466" class="stentry" rowspan="1" colspan="1">nsec, nanoseconds, usec, microseconds, msec, milliseconds,
                                       seconds
                                    </td>
                                    <td valign="top" headers="d54e4469" class="stentry" rowspan="1" colspan="1">nanoseconds</td>
                                    <td valign="top" headers="d54e4472" class="stentry" rowspan="1" colspan="1">Set basic unit of time for all rules. The argument of the
                                       switch is matched by using the longest prefix matching. Meaning that
                                       it is not necessary to write a whole word as the switch argument. It
                                       is similar to passing a ":time=&lt;unit&gt;" argument to every
                                       formatter, although the formatter uses more strict naming
                                       conventions. See "nsys analyze --help-formats column" for more
                                       detailed information on unit conversion.
                                    </td>
                                 </tr>
                              </table>
                           </div>
                        </div>
                     </div>
                     <div class="topic nested2" id="cli-analyze-command-switch-options"><a name="cli-analyze-command-switch-options" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cli-analyze-command-switch-options" name="cli-analyze-command-switch-options" shape="rect">1.3.9.&nbsp;CLI Analyze Command Switch Options</a></h3>
                        <div class="body">
                           <p class="p">
                              The <samp class="ph codeph"><span class="ph">nsys</span> analyze</samp> command generates
                              and outputs to the terminal a report using expert system rules on
                              existing results. Reports are generated from an SQLite export of a
                              .nsys-rep file. If a .nsys-rep file is specified, <span class="ph">Nsight Systems</span>
                              will look for an accompanying SQLite file and use it. If no SQLite export
                              file exists, one will be created.
                           </p>
                           <p class="p">After choosing the <samp class="ph codeph">analyze</samp> command switch, the
                              following options are available. Usage:
                           </p>
                           <p class="p"><samp class="ph codeph"><span class="ph">nsys</span> [global-options] analyze [options] [input-file]
                                 </samp></p>
                           <div class="p">
                              <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                                 <tr class="sthead">
                                    <th valign="bottom" align="left" id="d54e4739" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Short</th>
                                    <th valign="bottom" align="left" id="d54e4742" class="stentry" width="14.285714285714285%" rowspan="1" colspan="1">Long</th>
                                    <th valign="bottom" align="left" id="d54e4745" class="stentry" width="19.047619047619047%" rowspan="1" colspan="1">Possible Parameters</th>
                                    <th valign="bottom" align="left" id="d54e4748" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Default</th>
                                    <th valign="bottom" align="left" id="d54e4751" class="stentry" width="47.61904761904762%" rowspan="1" colspan="1">Switch Description</th>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4739" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4742" class="stentry" rowspan="1" colspan="1">--help</td>
                                    <td valign="top" headers="d54e4745" class="stentry" rowspan="1" colspan="1">&lt;tag&gt;</td>
                                    <td valign="top" headers="d54e4748" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e4751" class="stentry" rowspan="1" colspan="1">Print the help message. The option can take one optional
                                       argument that will be used as a tag. If a tag is provided, only options
                                       relevant to the tag will be printed.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4739" class="stentry" rowspan="1" colspan="1">-f</td>
                                    <td valign="top" headers="d54e4742" class="stentry" rowspan="1" colspan="1">--format</td>
                                    <td valign="top" headers="d54e4745" class="stentry" rowspan="1" colspan="1">column, table, csv, tsv, json, hdoc, htable, .</td>
                                    <td valign="top" headers="d54e4748" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4751" class="stentry" rowspan="1" colspan="1">Specify the output format of the corresponding report(s). The special name "."
                                       indicates the default format for the given output. The default format for console is
                                       column, while files and process outputs default to csv. This option may be used multiple
                                       times. Multiple formats may also be specified using a comma-separated list
                                       (&lt;name[:args...][,name[:args...]...]&gt;). See Report Scripts for options available with
                                       each format.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4739" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4742" class="stentry" rowspan="1" colspan="1">--force-export</td>
                                    <td valign="top" headers="d54e4745" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e4748" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e4751" class="stentry" rowspan="1" colspan="1">Force a re-export of the SQLite file from the specified
                                       .nsys-rep file, even if an SQLite file already exists.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4739" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4742" class="stentry" rowspan="1" colspan="1">--force-overwrite</td>
                                    <td valign="top" headers="d54e4745" class="stentry" rowspan="1" colspan="1">true, false</td>
                                    <td valign="top" headers="d54e4748" class="stentry" rowspan="1" colspan="1">false</td>
                                    <td valign="top" headers="d54e4751" class="stentry" rowspan="1" colspan="1">Overwrite any existing output files.</td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4739" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4742" class="stentry" rowspan="1" colspan="1">--help-formats</td>
                                    <td valign="top" headers="d54e4745" class="stentry" rowspan="1" colspan="1">&lt;format_name&gt;, ALL, [none]</td>
                                    <td valign="top" headers="d54e4748" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e4751" class="stentry" rowspan="1" colspan="1">With no argument, list a summary of the available output formats.
                                       If a format name is given, a more detailed explanation of the
                                       the format is displayed. If <samp class="ph codeph">ALL</samp> is given, a more detailed
                                       explanation of all available formats is displayed.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4739" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4742" class="stentry" rowspan="1" colspan="1">--help-rules</td>
                                    <td valign="top" headers="d54e4745" class="stentry" rowspan="1" colspan="1">&lt;report_name&gt;, ALL, [none]</td>
                                    <td valign="top" headers="d54e4748" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e4751" class="stentry" rowspan="1" colspan="1">With no argument, list available rules with a short
                                       description. If a rule name is given, a more detailed
                                       explanation of the rule is displayed.  If <samp class="ph codeph">ALL</samp> is
                                       given, a more detailed explanation of all available rules is
                                       displayed.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4739" class="stentry" rowspan="1" colspan="1">-o</td>
                                    <td valign="top" headers="d54e4742" class="stentry" rowspan="1" colspan="1">--output</td>
                                    <td valign="top" headers="d54e4745" class="stentry" rowspan="1" colspan="1">-, @&lt;command&gt;, &lt;basename&gt;, .</td>
                                    <td valign="top" headers="d54e4748" class="stentry" rowspan="1" colspan="1">-</td>
                                    <td valign="top" headers="d54e4751" class="stentry" rowspan="1" colspan="1">Specify the output mechanism for the corresponding rule(s).
                                       There are three output mechanisms: print to console, output to
                                       file, or output to command.
                                       
                                       This option may be used multiple times. Multiple outputs may also
                                       be specified using a comma-separated list.
                                       
                                       If the given output name is "-", the rule will be displayed
                                       on the console. If the output name starts with "@", the output
                                       designates a command to run. The command will be executed and
                                       the rule output will be piped into the command. Any other output
                                       is assumed to be the base path and name for a file.
                                       
                                       If a file basename is given, the filename used will be:
                                       
                                       &lt;basename&gt;_&lt;report&amp;args&gt;.&lt;output_format&gt;.
                                       
                                       The default base (including path) is the name of the SQLite file
                                       (as derived from the input file or --sqlite option), minus the
                                       extension. The output "." can be used to indicate the rule
                                       should be output to a file, and the default basename should be
                                       used. To write one or more rules to files using the default
                                       basename, use the option: "--output .".
                                       
                                       If the output starts with "@", the rule is output to the
                                       given command. The command is run, and the output of the rule
                                       is piped to the command's stdin (standard-input). The command's
                                       stdout and stderr remain attached to the console, so any output
                                       will be displayed directly to the console.
                                       
                                       Be aware there are some limitations in how the command string is
                                       parsed. No shell expansions (including *, ?, [], and ~) are
                                       supported. The command cannot be piped to another command, nor
                                       redirected to a file using shell syntax. The command and command
                                       arguments are split on whitespace, and no quotes (within the command
                                       syntax) are supported. For commands that require complex command
                                       line syntax, it is suggested that the command be put into a shell
                                       script file, and the script designated as the output command.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4739" class="stentry" rowspan="1" colspan="1">-q</td>
                                    <td valign="top" headers="d54e4742" class="stentry" rowspan="1" colspan="1">--quiet</td>
                                    <td valign="top" headers="d54e4745" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4748" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4751" class="stentry" rowspan="1" colspan="1">Do not display verbose messages, only display errors.</td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4739" class="stentry" rowspan="1" colspan="1">-r</td>
                                    <td valign="top" headers="d54e4742" class="stentry" rowspan="1" colspan="1">--rule</td>
                                    <td valign="top" headers="d54e4745" class="stentry" rowspan="1" colspan="1">cuda-async-memcpy, cuda-sync-memcpy, cuda-sync-memset,
                                       cuda-sync-api, gpu-starv, gpu-low-util, dx12-mem-op
                                    </td>
                                    <td valign="top" headers="d54e4748" class="stentry" rowspan="1" colspan="1">all</td>
                                    <td valign="top" headers="d54e4751" class="stentry" rowspan="1" colspan="1">Specify the rules(s) to execute, including any arguments.
                                       This option may be used multiple times. Multiple reports may also be
                                       specified using a comma-separated list. See <strong class="ph b">Expert Systems</strong>
                                       section and --help-rules switch for details on all rules.
                                       
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4739" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4742" class="stentry" rowspan="1" colspan="1">--sqlite</td>
                                    <td valign="top" headers="d54e4745" class="stentry" rowspan="1" colspan="1">&lt;file.sqlite&gt;</td>
                                    <td valign="top" headers="d54e4748" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4751" class="stentry" rowspan="1" colspan="1">Specify the SQLite export filename. If this file exists, it
                                       will be used. If this file doesn't exist (or if --force-export was
                                       given) this file will be created from the specified .nsys-rep file
                                       before report processing. This option cannot be used if the specified
                                       input file is also an SQLite file.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4739" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4742" class="stentry" rowspan="1" colspan="1">--timeunit</td>
                                    <td valign="top" headers="d54e4745" class="stentry" rowspan="1" colspan="1">nsec, nanoseconds, usec, microseconds, msec, milliseconds, seconds</td>
                                    <td valign="top" headers="d54e4748" class="stentry" rowspan="1" colspan="1">nanoseconds</td>
                                    <td valign="top" headers="d54e4751" class="stentry" rowspan="1" colspan="1">Set basic unit of time for all rules. The argument of the
                                       switch is matched by using the longest prefix matching. Meaning that
                                       it is not necessary to write a whole word as the switch argument. It
                                       is similar to passing a ":time=&lt;unit&gt;" argument to every
                                       formatter, although the formatter uses more strict naming
                                       conventions. See "nsys analyze --help-formats column" for more
                                       detailed information on unit conversion.
                                    </td>
                                 </tr>
                              </table>
                           </div>
                        </div>
                     </div>
                     <div class="topic nested2" id="cli-status-command-switch-options"><a name="cli-status-command-switch-options" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cli-status-command-switch-options" name="cli-status-command-switch-options" shape="rect">1.3.10.&nbsp;CLI Status Command Switch Options</a></h3>
                        <div class="body">
                           <p class="p">The <samp class="ph codeph"><span class="ph">nsys</span> status</samp> command returns the current state of
                              the CLI. After choosing the <samp class="ph codeph">status</samp> command switch, the following options
                              are available. Usage:
                           </p><pre xml:space="preserve"><span class="ph">nsys</span> [global-options] status [options]</pre><div class="p">
                              <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                                 <tr class="sthead">
                                    <th valign="bottom" align="left" id="d54e4987" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Short</th>
                                    <th valign="bottom" align="left" id="d54e4990" class="stentry" width="14.285714285714285%" rowspan="1" colspan="1">Long</th>
                                    <th valign="bottom" align="left" id="d54e4993" class="stentry" width="19.047619047619047%" rowspan="1" colspan="1">Possible Parameters</th>
                                    <th valign="bottom" align="left" id="d54e4996" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Default</th>
                                    <th valign="bottom" align="left" id="d54e4999" class="stentry" width="47.61904761904762%" rowspan="1" colspan="1">Switch Description</th>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4987" class="stentry" rowspan="1" colspan="1">-e</td>
                                    <td valign="top" headers="d54e4990" class="stentry" rowspan="1" colspan="1">--environment</td>
                                    <td valign="top" headers="d54e4993" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4996" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4999" class="stentry" rowspan="1" colspan="1">Returns information about the system regarding suitability of
                                       the profiling environment.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4987" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4990" class="stentry" rowspan="1" colspan="1">--help</td>
                                    <td valign="top" headers="d54e4993" class="stentry" rowspan="1" colspan="1">&lt;tag&gt;</td>
                                    <td valign="top" headers="d54e4996" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e4999" class="stentry" rowspan="1" colspan="1">Print the help message. The option can take one optional
                                       argument that will be used as a tag. If a tag is provided, only options
                                       relevant to the tag will be printed.
                                    </td>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e4987" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                    <td valign="top" headers="d54e4990" class="stentry" rowspan="1" colspan="1">--session</td>
                                    <td valign="top" headers="d54e4993" class="stentry" rowspan="1" colspan="1">session identifier</td>
                                    <td valign="top" headers="d54e4996" class="stentry" rowspan="1" colspan="1">none</td>
                                    <td valign="top" headers="d54e4999" class="stentry" rowspan="1" colspan="1">Print the status of the indicated session. The option argument
                                       must represent a valid session name or ID as reported by
                                       <samp class="ph codeph"><span class="ph">nsys</span>sessions list</samp>. Any
                                       <samp class="ph codeph">%q{ENV_VAR}</samp> pattern will be substituted with the
                                       value of the environment variable. Any <samp class="ph codeph">%h</samp> pattern
                                       will be substituted with the hostname of the system. Any <samp class="ph codeph">%%</samp>
                                       pattern will be substituted with <samp class="ph codeph">%</samp>.
                                    </td>
                                 </tr>
                              </table>
                           </div>
                        </div>
                     </div>
                     <div class="topic nested2" id="cli-sessions-command-switch-subcommands"><a name="cli-sessions-command-switch-subcommands" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cli-sessions-command-switch-subcommands" name="cli-sessions-command-switch-subcommands" shape="rect">1.3.11.&nbsp;CLI Sessions Command Switch Subcommands</a></h3>
                        <div class="body">
                           <p class="p">After choosing the <samp class="ph codeph">sessions</samp> command switch, the
                              following subcommands are available. Usage:
                           </p><pre xml:space="preserve"><span class="ph">nsys</span> [global-options] sessions [subcommand]</pre><div class="p">
                              <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                                 <tr class="sthead">
                                    <th valign="bottom" align="left" id="d54e5098" class="stentry" width="16.666666666666668%" rowspan="1" colspan="1">Subcommand</th>
                                    <th valign="bottom" align="left" id="d54e5101" class="stentry" width="83.33333333333334%" rowspan="1" colspan="1">Description</th>
                                 </tr>
                                 <tr class="strow">
                                    <td valign="top" headers="d54e5098" class="stentry" rowspan="1" colspan="1">list</td>
                                    <td valign="top" headers="d54e5101" class="stentry" rowspan="1" colspan="1">List all active sessions including ID, name, and state information</td>
                                 </tr>
                              </table>
                           </div>
                        </div>
                        <div class="topic nested3" id="cli-sessions-list-command-switch-subcommands"><a name="cli-sessions-list-command-switch-subcommands" shape="rect">
                              <!-- --></a><h3 class="title topictitle2"><a href="#cli-sessions-list-command-switch-subcommands" name="cli-sessions-list-command-switch-subcommands" shape="rect">1.3.11.1.&nbsp;CLI Sessions List Command Switch Options</a></h3>
                           <div class="body">
                              <p class="p">After choosing the <samp class="ph codeph">sessions list</samp> command switch, the
                                 following options are available. Usage:
                              </p><pre xml:space="preserve"><span class="ph">nsys</span> [global-options] sessions list [options]</pre><div class="p">
                                 <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                                    <tr class="sthead">
                                       <th valign="bottom" align="left" id="d54e5142" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Short</th>
                                       <th valign="bottom" align="left" id="d54e5145" class="stentry" width="14.285714285714285%" rowspan="1" colspan="1">Long</th>
                                       <th valign="bottom" align="left" id="d54e5148" class="stentry" width="19.047619047619047%" rowspan="1" colspan="1">Possible Parameters</th>
                                       <th valign="bottom" align="left" id="d54e5151" class="stentry" width="9.523809523809524%" rowspan="1" colspan="1">Default</th>
                                       <th valign="bottom" align="left" id="d54e5154" class="stentry" width="47.61904761904762%" rowspan="1" colspan="1">Switch Description</th>
                                    </tr>
                                    <tr class="strow">
                                       <td valign="top" headers="d54e5142" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                       <td valign="top" headers="d54e5145" class="stentry" rowspan="1" colspan="1">--help</td>
                                       <td valign="top" headers="d54e5148" class="stentry" rowspan="1" colspan="1">&lt;tag&gt;</td>
                                       <td valign="top" headers="d54e5151" class="stentry" rowspan="1" colspan="1">none</td>
                                       <td valign="top" headers="d54e5154" class="stentry" rowspan="1" colspan="1">Print the help message. The option can take one optional
                                          argument that will be used as a tag. If a tag is provided, only options
                                          relevant to the tag will be printed.
                                       </td>
                                    </tr>
                                    <tr class="strow">
                                       <td valign="top" headers="d54e5142" class="stentry" rowspan="1" colspan="1">-p</td>
                                       <td valign="top" headers="d54e5145" class="stentry" rowspan="1" colspan="1">--show-header</td>
                                       <td valign="top" headers="d54e5148" class="stentry" rowspan="1" colspan="1">true, false</td>
                                       <td valign="top" headers="d54e5151" class="stentry" rowspan="1" colspan="1">true</td>
                                       <td valign="top" headers="d54e5154" class="stentry" rowspan="1" colspan="1">Controls whether a header should appear in the output.</td>
                                    </tr>
                                 </table>
                              </div>
                           </div>
                        </div>
                     </div>
                  </div>
                  <div class="topic nested1" id="example-single-command-lines"><a name="example-single-command-lines" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#example-single-command-lines" name="example-single-command-lines" shape="rect">1.4.&nbsp;Example Single Command Lines</a></h3>
                     <div class="body">
                        <p class="p"><strong class="ph b">Version Information</strong></p><pre xml:space="preserve"><span class="ph">nsys</span> -v</pre><p class="p">Effect: Prints tool version information to the screen.</p>
                        <p class="p"><strong class="ph b">Run with elevated privilege</strong></p><pre xml:space="preserve">sudo <span class="ph">nsys</span> profile &lt;app&gt;</pre><p class="p">Effect: <span class="ph">Nsight Systems</span> CLI (and target application) will run
                           with elevated privilege. This is necessary for some features, such as FTrace
                           or system-wide CPU sampling. If you don't want the target application to be
                           elevated, use `--run-as` option.
                        </p>
                        <p class="p"><strong class="ph b">Default analysis run</strong></p><pre xml:space="preserve"><span class="ph">nsys</span> profile &lt;application&gt;
    [application-arguments]</pre><p class="p">Effect: Launch the application using the given arguments. Start collecting
                           immediately and end collection when the application stops. Trace CUDA,
                           OpenGL, NVTX, and OS runtime libraries APIs. Collect CPU sampling information
                           and thread scheduling information. With <span class="ph">Nsight Systems Embedded Platforms Edition</span> this will
                           	only analysis the single process. With <span class="ph">Nsight Systems Workstation Edition</span> this
                           	will trace the process tree. Generate the report#.nsys-rep file in the default
                           	location, incrementing the report number if needed to avoid overwriting any
                           	existing output files.
                        </p>
                        <p class="p"><strong class="ph b">Limited trace only run</strong></p><pre xml:space="preserve"><span class="ph">nsys</span> profile --trace=cuda,nvtx -d 20
    --sample=none --cpuctxsw=none -o my_test &lt;application&gt;
    [application-arguments]</pre><p class="p">Effect: Launch the application using the given arguments. Start collecting
                           immediately and end collection after 20 seconds or when the application ends.
                           Trace CUDA and NVTX APIs. Do not collect CPU sampling information or thread
                           scheduling information. Profile any child processes. Generate the output file
                           as my_test.nsys-rep in the current working directory.
                        </p>
                        <p class="p"><strong class="ph b">Delayed start run</strong></p><pre xml:space="preserve"><span class="ph">nsys</span> profile -e TEST_ONLY=0 -y 20
    &lt;application&gt; [application-arguments]</pre><p class="p">Effect: Set environment variable TEST_ONLY=0. Launch the application
                           using the given arguments. Start collecting after 20 seconds and end
                           collection at application exit. Trace CUDA, OpenGL, NVTX, and OS runtime
                           libraries APIs. Collect CPU sampling and thread schedule information. Profile
                           any child processes. Generate the report#.nsys-rep file in the default location,
                           incrementing if needed to avoid overwriting any existing output files.
                        </p>
                        <p class="p"><strong class="ph b">Collect ftrace events</strong></p><pre xml:space="preserve"><span class="ph">nsys</span> profile --ftrace=drm/drm_vblank_event
    -d 20</pre><p class="p">Effect: Collect ftrace <samp class="ph codeph">drm_vblank_event</samp> events for 20
                           seconds. Generate the report#.nsys-rep file in the current working directory.
                           Note that ftrace event collection requires running as root. To get a list of
                           ftrace events available from the kernel, run the following:
                        </p><pre xml:space="preserve">sudo cat /sys/kernel/debug/tracing/available_events</pre><p class="p"><strong class="ph b">Run GPU metric sampling on one TU10x</strong></p><pre xml:space="preserve"><span class="ph">nsys</span> profile --gpu-metrics-device=0
	--gpu-metrics-set=tu10x-gfxt &lt;application&gt;</pre><p class="p">Effect: Launch application. Collect default options and GPU metrics for
                           	the first GPU (a TU10x), using the tu10x-gfxt metric set at the default
                           	frequency (10 kHz). Profile any child processes. Generate the report#.nsys-rep
                           	file in the default location, incrementing if needed to avoid overwriting
                           	any existing output files.
                        </p>
                        <p class="p"><strong class="ph b">Run GPU metric sampling on all GPUs at a set frequency</strong></p><pre xml:space="preserve"><span class="ph">nsys</span> profile --gpu-metrics-device=all
	--gpu-metrics-frequency=20000 &lt;application&gt;</pre><p class="p">Effect: Launch application. Collect default options and GPU metrics for
                           	all available GPUs using the first suitable metric set for each and sampling
                           	at 20 kHz. Profile any child processes. Generate the report#.nsys-rep
                           	file in the default location, incrementing if needed to avoid overwriting
                           	any existing output files.
                        </p>
                        <p class="p"><strong class="ph b">Collect custom ETW trace using configuration file</strong></p><pre xml:space="preserve"><span class="ph">nsys</span> profile --etw-provider=file.JSON</pre><p class="p">Effect: Configure custom ETW collectors using the contents of file.JSON.
                           	Collect data for 20 seconds. Generate the report#.nsys-rep file in the current
                           	working directory.
                        </p>
                        <div class="p">
                           	A template JSON configuration file is located at in the <span class="ph">Nsight Systems</span>
                           	installation directory as \target-windows-x64\etw_providers_template.json.
                           	This path will show up automatically if you call
                           	<pre xml:space="preserve"><span class="ph">nsys</span> profile --help</pre></div>
                        <div class="p">
                           	The <strong class="ph b">level</strong> attribute can only be set to one of the following:
                           	
                           <ul class="ul">
                              <li class="li">TRACE_LEVEL_CRITICAL</li>
                              <li class="li">TRACE_LEVEL_ERROR</li>
                              <li class="li">TRACE_LEVEL_WARNING</li>
                              <li class="li">TRACE_LEVEL_INFORMATION</li>
                              <li class="li">TRACE_LEVEL_VERBOSE</li>
                           </ul>
                        </div>
                        <div class="p">
                           	The <strong class="ph b">flags</strong> attribute can only be set to one or more of the
                           	following:
                           	
                           <ul class="ul">
                              <li class="li">EVENT_TRACE_FLAG_ALPC</li>
                              <li class="li">EVENT_TRACE_FLAG_CSWITCH</li>
                              <li class="li">EVENT_TRACE_FLAG_DBGPRINT</li>
                              <li class="li">EVENT_TRACE_FLAG_DISK_FILE_IO</li>
                              <li class="li">EVENT_TRACE_FLAG_DISK_IO</li>
                              <li class="li">EVENT_TRACE_FLAG_DISK_IO_INIT</li>
                              <li class="li">EVENT_TRACE_FLAG_DISPATCHER</li>
                              <li class="li">EVENT_TRACE_FLAG_DPC</li>
                              <li class="li">EVENT_TRACE_FLAG_DRIVER</li>
                              <li class="li">EVENT_TRACE_FLAG_FILE_IO</li>
                              <li class="li">EVENT_TRACE_FLAG_FILE_IO_INIT</li>
                              <li class="li">EVENT_TRACE_FLAG_IMAGE_LOAD</li>
                              <li class="li">EVENT_TRACE_FLAG_INTERRUPT</li>
                              <li class="li">EVENT_TRACE_FLAG_JOB</li>
                              <li class="li">EVENT_TRACE_FLAG_MEMORY_HARD_FAULTS</li>
                              <li class="li">EVENT_TRACE_FLAG_MEMORY_PAGE_FAULTS</li>
                              <li class="li">EVENT_TRACE_FLAG_NETWORK_TCPIP</li>
                              <li class="li">EVENT_TRACE_FLAG_NO_SYSCONFIG</li>
                              <li class="li">EVENT_TRACE_FLAG_PROCESS</li>
                              <li class="li">EVENT_TRACE_FLAG_PROCESS_COUNTERS</li>
                              <li class="li">EVENT_TRACE_FLAG_PROFILE</li>
                              <li class="li">EVENT_TRACE_FLAG_REGISTRY</li>
                              <li class="li">EVENT_TRACE_FLAG_SPLIT_IO</li>
                              <li class="li">EVENT_TRACE_FLAG_SYSTEMCALL</li>
                              <li class="li">EVENT_TRACE_FLAG_THREAD</li>
                              <li class="li">EVENT_TRACE_FLAG_VAMAP</li>
                              <li class="li">EVENT_TRACE_FLAG_VIRTUAL_ALLOC</li>
                           </ul>
                        </div>
                        <p class="p"><strong class="ph b">Typical case: profile a Python script that uses CUDA</strong></p><pre xml:space="preserve"><span class="ph">nsys</span> profile --trace=cuda,cudnn,cublas,osrt,nvtx
    --delay=60 python my_dnn_script.py</pre><p class="p">Effect: Launch a Python script and start profiling it 60 seconds after
                           the launch, tracing CUDA, cuDNN, cuBLAS, OS runtime APIs, and NVTX as well
                           as collecting thread schedule information.
                        </p>
                        <p class="p"><strong class="ph b">Typical case: profile an app that uses Vulkan</strong></p><pre xml:space="preserve"><span class="ph">nsys</span> profile --trace=vulkan,osrt,nvtx
    --delay=60 ./myapp</pre><p class="p">Effect: Launch an app and start profiling it 60 seconds after the launch,
                           tracing Vulkan, OS runtime APIs, and NVTX as well as collecting CPU sampling
                           and thread schedule information.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="example-interactive-cli-command-sequences"><a name="example-interactive-cli-command-sequences" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#example-interactive-cli-command-sequences" name="example-interactive-cli-command-sequences" shape="rect">1.5.&nbsp;Example Interactive CLI Command Sequences</a></h3>
                     <div class="body">
                        <p class="p"><strong class="ph b">Collect from beginning of application, end manually</strong></p><pre xml:space="preserve"><span class="ph">nsys</span> start --stop-on-exit=false
<span class="ph">nsys</span> launch --trace=cuda,nvtx --sample=none &lt;application&gt; [application-arguments]
<span class="ph">nsys</span> stop</pre><p class="p">Effect: Create interactive CLI process and set it up to begin collecting
                           as soon as an application is launched. Launch the application, set up to
                           allow tracing of CUDA and NVTX as well as collection of thread schedule
                           information. Stop only when explicitly requested. Generate the
                           report#.nsys-rep in the default location.
                        </p>
                        <blockquote class="lq">
                           <div class="p">
                              <div class="tablenoborder">
                                 <table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
                                    <tbody class="tbody">
                                       <tr class="row">
                                          <td class="entry nvidiagreen" align="center" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                             <p class="p">&nbsp;&nbsp;<strong class="ph b">Note:</strong>&nbsp;&nbsp;
                                             </p>
                                          </td>
                                          <td class="entry notebox" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                             <p class="p">If you start a collection and fail to stop the collection (or if
                                                you are allowing it to stop on exit, and the application runs for too
                                                long) your system’s storage space may be filled with collected data
                                                causing significant issues for the system. <span class="ph">Nsight Systems</span>
                                                will collect a different amount of data/sec depending on options, but in
                                                general <span class="ph">Nsight Systems</span> does not support runs of more than
                                                5 minutes duration.
                                             </p>
                                          </td>
                                       </tr>
                                    </tbody>
                                 </table>
                              </div>
                           </div>
                        </blockquote>
                        <p class="p"><strong class="ph b">Run application, begin collection manually, run until process ends
                              </strong></p><pre xml:space="preserve"><span class="ph">nsys</span> launch -w true &lt;application&gt; [application-arguments]
<span class="ph">nsys</span> start</pre><p class="p">Effect: Create interactive CLI and launch an application set up for default
                           analysis. Send application output to the terminal. No data is collected
                           until you manually start collection at area of interest. Profile until
                           the application ends. Generate the report#.nsys-rep in the default location.
                           
                        </p>
                        <blockquote class="lq">
                           <div class="p">
                              <div class="tablenoborder">
                                 <table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
                                    <tbody class="tbody">
                                       <tr class="row">
                                          <td class="entry nvidiagreen" align="center" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                             <p class="p">&nbsp;&nbsp;<strong class="ph b">Note:</strong>&nbsp;&nbsp;
                                             </p>
                                          </td>
                                          <td class="entry notebox" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                             <p class="p">If you launch an application and that application and any
                                                descendants exit before start is called <span class="ph">Nsight Systems</span>
                                                will create a fully formed .nsys-rep file containing no data.&nbsp;&nbsp;
                                             </p>
                                          </td>
                                       </tr>
                                    </tbody>
                                 </table>
                              </div>
                           </div>
                        </blockquote>
                        <p class="p"><strong class="ph b">Run application, start/stop collection using cudaProfilerStart/Stop
                              </strong></p><pre xml:space="preserve"><span class="ph">nsys</span> start -c cudaProfileApi
<span class="ph">nsys</span> launch -w true &lt;application&gt; [application-arguments]</pre><p class="p">Effect: Create interactive CLI process and set it up to begin collecting
                           as soon as a cudaProfileStart() is detected. Launch application for
                           default analysis, sending application output to the terminal. Stop
                           collection at next call to cudaProfilerStop, when the user calls 
                           <samp class="ph codeph"><span class="ph">nsys</span> stop</samp>, or when the root process
                           terminates. Generate the report#.nsys-rep in the default location.
                           
                        </p>
                        <blockquote class="lq">
                           <div class="p">
                              <div class="tablenoborder">
                                 <table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
                                    <tbody class="tbody">
                                       <tr class="row">
                                          <td class="entry nvidiagreen" align="center" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                             <p class="p">&nbsp;&nbsp;<strong class="ph b">Note:</strong>&nbsp;&nbsp;
                                             </p>
                                          </td>
                                          <td class="entry notebox" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                             <p class="p">If you call <samp class="ph codeph"><span class="ph">nsys</span> launch</samp> before 
                                                <samp class="ph codeph"><span class="ph">nsys</span> start -c cudaProfilerApi</samp> and the
                                                code contains a large number of short duration cudaProfilerStart/Stop pairs,
                                                <span class="ph">Nsight Systems</span> may be unable to process them correctly,
                                                causing a fault. This will be corrected in a future version.&nbsp;&nbsp;
                                             </p>
                                          </td>
                                       </tr>
                                    </tbody>
                                 </table>
                              </div>
                           </div>
                        </blockquote>
                        <blockquote class="lq">
                           <div class="p">
                              <div class="tablenoborder">
                                 <table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
                                    <tbody class="tbody">
                                       <tr class="row">
                                          <td class="entry nvidiagreen" align="center" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                             <p class="p">&nbsp;&nbsp;<strong class="ph b">Note:</strong>&nbsp;&nbsp;
                                             </p>
                                          </td>
                                          <td class="entry notebox" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                             <p class="p">The <span class="ph">Nsight Systems</span> CLI does not support multiple calls
                                                to the cudaProfilerStart/Stop API at this time.  &nbsp;&nbsp;
                                             </p>
                                          </td>
                                       </tr>
                                    </tbody>
                                 </table>
                              </div>
                           </div>
                        </blockquote>
                        <p class="p"><strong class="ph b">Run application, start/stop collection using NVTX</strong></p><pre xml:space="preserve"><span class="ph">nsys</span> start -c nvtx
<span class="ph">nsys</span> launch -w true -p MESSAGE@DOMAIN &lt;application&gt; [application-arguments]</pre><p class="p">Effect: Create interactive CLI process and set it up to begin collecting
                           as soon as an NVTX range with given message in given domain (capture
                           range) is opened. Launch application for default analysis, sending
                           application output to the terminal. Stop collection when all capture
                           ranges are closed, when the user calls <samp class="ph codeph"><span class="ph">nsys</span> stop</samp>, 
                           or when the root process terminates. Generate the report#.nsys-rep in the
                           default location.
                        </p>
                        <blockquote class="lq">
                           <div class="p">
                              <div class="tablenoborder">
                                 <table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
                                    <tbody class="tbody">
                                       <tr class="row">
                                          <td class="entry nvidiagreen" align="center" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                             <p class="p">&nbsp;&nbsp;<strong class="ph b">Note:</strong>&nbsp;&nbsp;
                                             </p>
                                          </td>
                                          <td class="entry notebox" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                             <p class="p">The <span class="ph">Nsight Systems</span> CLI only triggers the profiling
                                                session for the first capture range.&nbsp;&nbsp;
                                             </p>
                                          </td>
                                       </tr>
                                    </tbody>
                                 </table>
                              </div>
                           </div>
                        </blockquote>
                        <p class="p">NVTX capture range can be specified:</p>
                        <ul class="ul">
                           <li class="li">
                              <p class="p">Message@Domain: All ranges with given message in given domain are
                                 capture ranges. For example:
                              </p><pre xml:space="preserve"><span class="ph">nsys</span> launch -w true -p profiler@service ./app</pre><p class="p">This would make the profiling start when the first range with message
                                 "profiler" is opened in domain "service".
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">Message@*: All ranges with given message in all domains are capture
                                 ranges. For example:
                              </p><pre xml:space="preserve"><span class="ph">nsys</span> launch -w true -p profiler@* ./app</pre><p class="p">This would make the profiling start when the first range with message
                                 "profiler" is opened in any domain.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">Message: All ranges with given message in default domain are capture
                                 ranges. For example:
                              </p><pre xml:space="preserve"><span class="ph">nsys</span> launch -w true -p profiler ./app</pre><p class="p">This would make the profiling start when the first range with message "profiler"
                                 is opened in the default domain.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">By default only messages, provided by NVTX registered strings are
                                 considered to avoid additional overhead. To enable non-registered strings
                                 check please launch your application with
                                 <samp class="ph codeph">NSYS_NVTX_PROFILER_REGISTER_ONLY=0</samp> environment:
                              </p><pre xml:space="preserve"><span class="ph">nsys</span> launch -w true -p profiler@service -e NSYS_NVTX_PROFILER_REGISTER_ONLY=0 ./app</pre></li>
                        </ul>
                        <p class="p"><strong class="ph b">Run application, start/stop collection multiple times</strong></p>
                        <p class="p">The interactive CLI supports multiple sequential collections per launch.</p><pre xml:space="preserve"><span class="ph">nsys</span> launch &lt;application&gt; [application-arguments]
<span class="ph">nsys</span> start
<span class="ph">nsys</span> stop
<span class="ph">nsys</span> start
<span class="ph">nsys</span> stop
<span class="ph">nsys</span> shutdown --kill sigkill</pre><p class="p">Effect: Create interactive CLI and launch an application set up for
                           default analysis. Send application output to the terminal. No data is
                           collected until the start command is executed. Collect data from start
                           until stop requested, generate report#.qstrm in the current working
                           directory. Collect data from second start until the second stop request,
                           generate report#.nsys-rep (incremented by one) in the current working
                           directory. Shutdown the interactive CLI and send sigkill to the target
                           application's process group.
                           
                        </p>
                        <blockquote class="lq">
                           <div class="p">
                              <div class="tablenoborder">
                                 <table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
                                    <tbody class="tbody">
                                       <tr class="row">
                                          <td class="entry nvidiagreen" align="center" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                             <p class="p">&nbsp;&nbsp;<strong class="ph b">Note:</strong>&nbsp;&nbsp;
                                             </p>
                                          </td>
                                          <td class="entry notebox" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                             <p class="p">Calling <samp class="ph codeph"><span class="ph">nsys</span> cancel</samp> after 
                                                <samp class="ph codeph"><span class="ph">nsys</span> start</samp> will cancel the collection
                                                without generating a report.&nbsp;&nbsp;
                                             </p>
                                          </td>
                                       </tr>
                                    </tbody>
                                 </table>
                              </div>
                           </div>
                        </blockquote>
                     </div>
                  </div>
                  <div class="topic nested1" id="example-nsys-stats"><a name="example-nsys-stats" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#example-nsys-stats" name="example-nsys-stats" shape="rect">1.6.&nbsp;Example Stats Command Sequences</a></h3>
                     <div class="body">
                        <p class="p"><strong class="ph b">Display default statistics</strong></p>
                        <p class="p"><samp class="ph codeph"><span class="ph">nsys</span> stats report1.nsys-rep</samp></p>
                        <p class="p">Effect: Export an SQLite file named report1.sqlite from report1.nsys-rep
                           (assuming it does not already exist). Print the default reports in column
                           format to the console.
                        </p>
                        <p class="p">
                           Note: The following two command sequences should present very similar
                           information:
                        </p>
                        <p class="p"><samp class="ph codeph"><span class="ph">nsys</span> profile --stats=true &lt;application&gt;
                              </samp></p>
                        <p class="p">or</p>
                        <p class="p"><samp class="ph codeph"><span class="ph">nsys</span> profile &lt;application&gt;</samp></p>
                        <p class="p"><samp class="ph codeph"><span class="ph">nsys</span> stats report1.nsys-rep</samp></p>
                        <p class="p"><strong class="ph b">Display specific data from a report</strong></p>
                        <p class="p"><samp class="ph codeph"><span class="ph">nsys</span> stats --report gputrace report1.nsys-rep</samp></p>
                        <p class="p">Effect: Export an SQLite file named report1.sqlite from report1.nsys-rep
                           (assuming it does not already exist). Print the report generated by the
                           <samp class="ph codeph">gputrace</samp> script to the console in column format.
                        </p>
                        <p class="p"><strong class="ph b">Generate multiple reports, in multiple formats, output multiple
                              places</strong></p>
                        <p class="p"><samp class="ph codeph"><span class="ph">nsys</span> stats --report gputrace --report gpukernsum
                              --report cudaapisum --format csv,column --output .,- report1.nsys-rep
                              </samp></p>
                        <p class="p">Effect: Export an SQLite file named report1.sqlite from report1.nsys-rep
                           (assuming it does not already exist). Generate three reports. The first,
                           the <samp class="ph codeph">gputrace</samp> report, will be output to the file
                           report1_gputrace.csv in CSV format. The other two reports,
                           <samp class="ph codeph">gpukernsum</samp> and <samp class="ph codeph">cudaapisum</samp>, will be output
                           to the console as columns of data. Although three reports were given, only
                           two formats and outputs are given. To reconcile this, both the list of
                           formats and outputs is expanded to match the list of reports by repeating
                           the last element.
                        </p>
                        <p class="p"><strong class="ph b">Submit report data to a command</strong></p>
                        <p class="p"><samp class="ph codeph"><span class="ph">nsys</span> stats --report cudaapisum --format table \
                              --output @“grep -E (-|Name|cudaFree” test.sqlite
                              </samp></p>
                        <p class="p">Effect: Open test.sqlite and run the <samp class="ph codeph">
                              cudaapisum</samp> script on that file. Generate table data and feed that
                           into the command <samp class="ph codeph">grep -E (-|Name|cudaFree)</samp>. The grep
                           command will filter out everything but the header, formatting, and the
                           <samp class="ph codeph">cudaFree</samp> data, and display the results to the console.
                        </p>
                        <p class="p">
                           Note: When the output name starts with @, it is defined as a command. The
                           command is run, and the output of the report is piped to the command's
                           stdin (standard-input). The command's stdout and stderr remain attached to the
                           console, so any output will be displayed directly to the console.
                        </p>
                        <p class="p">
                           Be aware there are some limitations in how the command string is parsed. No
                           shell expansions (including *, ?, [], and ~) are supported. The command cannot
                           be piped to another command, nor redirected to a file using shell syntax.
                           The command and command arguments are split on whitespace, and no quotes
                           (within the command syntax) are supported. For commands that require
                           complex command line syntax, it is suggested that the command be put into a
                           shell script file, and the script designated as the output command
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="example-output-stats-option"><a name="example-output-stats-option" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#example-output-stats-option" name="example-output-stats-option" shape="rect">1.7.&nbsp;Example Output from --stats Option</a></h3>
                     <div class="body">
                        <p class="p"> The <samp class="ph codeph"><span class="ph">nsys</span> stats</samp> command can be
                           used post analysis to generate specific or personalized reports. For
                           a default fixed set of summary statistics to be automatically generated,
                           you can use the --stats option with the <samp class="ph codeph"><span class="ph">nsys</span>
                              profile</samp> or <samp class="ph codeph"><span class="ph">nsys</span> start</samp>
                           command to generate a fixed set of useful summary statistics.
                        </p>
                        <p class="p">If your run traces CUDA, these include CUDA API, Kernel, and Memory
                           Operation statistics:
                           
                        </p>
                        <p class="p"><img class="image" src="graphics/3407-profile-cuda-summary.png" height="-1" width="-1" alt="CUDA Statistics"></img></p>
                        <p class="p">If your run traces OS runtime events or NVTX push-pop ranges:</p>
                        <p class="p"><img class="image" src="graphics/3407-profile-osrt-nvtx-summary.png" height="-1" width="-1" alt="OS runtime and NVTX Statistics"></img></p>
                        <p class="p">If your run traces graphics debug markers these include DX11 debug markers, DX12
                           debug markers, Vulkan debug markers or KHR debug markers:
                        </p>
                        <p class="p"><img class="image" src="graphics/ue4_infiltrator_vulkan_markers-summary.png" height="-1" width="-1" alt="Graphics Vulkan debug markers Statistics"></img></p>
                        <p class="p">Recipes for these statistics as well as documentation on how to create
                           your own metrics will be available in a future version of the tool.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="importing-qdstrm-files"><a name="importing-qdstrm-files" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#importing-qdstrm-files" name="importing-qdstrm-files" shape="rect">1.8.&nbsp;Importing and Viewing Command Line Results Files</a></h3>
                     <div class="body">
                        <p class="p">The CLI generates a .qdstrm file. The .qdstrm file is an intermediate
                           result file, not intended for multiple imports. It needs to be processed,
                           either by importing it into the GUI or by using the standalone
                           QdstrmImporter to generate an optimized .nsys-rep file. Use this .nsys-rep file
                           when re-opening the result on the same machine, opening the result on a
                           different machine, or sharing results with teammates.
                        </p>
                        <p class="p">This version of <span class="ph">Nsight Systems</span> will attempt to
                           automatically convert the .qdstrm file to a .nsys-rep file with the same name
                           after the run finishes if the required libraries are available. The ability
                           to turn off auto-conversion will be added in a later version.
                        </p>
                        <p class="p"><strong class="ph b">Import Into the GUI</strong></p>
                        <p class="p">The CLI and host GUI versions must match to import a .qdstrm file
                           successfully. The host GUI is backward compatible only with .nsys-rep files.
                           
                        </p>
                        <p class="p">Copy the .qdstrm file you are interested in viewing to a system
                           where the <span class="ph">Nsight Systems</span> host GUI is installed. Launch the
                           <span class="ph">Nsight Systems</span> GUI. Select <strong class="ph b">File-&gt;Import...</strong> and choose
                           the .qdstrm file you wish to open.
                           
                        </p>
                        <blockquote class="lq">
                           <p class="p"><img class="image" src="graphics/import_qdstrm.png" alt="Import qdstrm"></img></p>
                        </blockquote>
                        <p class="p">The import of really large, multi-gigabyte, .qdstrm
                           files may take up all of the memory on the host computer and lock up the
                           system. This will be fixed in a later version.
                           
                        </p>
                        <p class="p"><strong class="ph b">Importing Windows ETL files</strong></p>
                        <p class="p">For Windows targets, ETL files captured with Xperf or the
                           <samp class="ph codeph">log.cmd</samp> command supplied with GPUView in the Windows
                           Performance Toolkit can be imported to create reports as if they were
                           captured with <span class="ph">Nsight Systems</span>'s "WDDM trace" and "Custom ETW
                           trace" features. Simply choose the .etl file from the Import dialog to
                           convert it to a .nsys-rep file.
                           
                        </p>
                        <p class="p"><strong class="ph b">Create .nsys-rep Using QdstrmImporter</strong></p>
                        <p class="p">The CLI and QdstrmImporter versions must match to convert a .qdstrm
                           file into a .nsys-rep file. This .nsys-rep file can then be opened in the same
                           version or more recent versions of the GUI.
                        </p>
                        <p class="p">To run QdstrmImporter on the host system, find the QdstrmImporter binary
                           in the Host-x86_64 directory in your installation. QdstrmImporter is
                           available for all host platforms. See options below.
                        </p>
                        <p class="p">To run QdstrmImporter on the target system, copy the Linux Host-x86_64
                           directory to the target Linux system or install <span class="ph">Nsight Systems</span>
                           for Linux host directly on the target. The Windows or macOS host
                           QdstrmImporter will not work on a Linux Target. See options below.
                        </p>
                        <div class="p">
                           <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                              <tr class="sthead">
                                 <th valign="bottom" align="left" id="d54e6179" class="stentry" width="10%" rowspan="1" colspan="1">Short</th>
                                 <th valign="bottom" align="left" id="d54e6182" class="stentry" width="20%" rowspan="1" colspan="1">Long</th>
                                 <th valign="bottom" align="left" id="d54e6185" class="stentry" width="20%" rowspan="1" colspan="1">Parameter</th>
                                 <th valign="bottom" align="left" id="d54e6188" class="stentry" width="50%" rowspan="1" colspan="1">Description</th>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e6179" class="stentry" rowspan="1" colspan="1">-h</td>
                                 <td valign="top" headers="d54e6182" class="stentry" rowspan="1" colspan="1">--help</td>
                                 <td valign="top" headers="d54e6185" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                 <td valign="top" headers="d54e6188" class="stentry" rowspan="1" colspan="1">Help message providing information about available options and
                                    their parameters.
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e6179" class="stentry" rowspan="1" colspan="1">-v</td>
                                 <td valign="top" headers="d54e6182" class="stentry" rowspan="1" colspan="1">--version</td>
                                 <td valign="top" headers="d54e6185" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                 <td valign="top" headers="d54e6188" class="stentry" rowspan="1" colspan="1">Output QdstrmImporter version information</td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e6179" class="stentry" rowspan="1" colspan="1">-i</td>
                                 <td valign="top" headers="d54e6182" class="stentry" rowspan="1" colspan="1">--input-file</td>
                                 <td valign="top" headers="d54e6185" class="stentry" rowspan="1" colspan="1">filename or path</td>
                                 <td valign="top" headers="d54e6188" class="stentry" rowspan="1" colspan="1">Import .qdstrm file from this location.</td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e6179" class="stentry" rowspan="1" colspan="1">-o</td>
                                 <td valign="top" headers="d54e6182" class="stentry" rowspan="1" colspan="1">--output-file</td>
                                 <td valign="top" headers="d54e6185" class="stentry" rowspan="1" colspan="1">filename or path</td>
                                 <td valign="top" headers="d54e6188" class="stentry" rowspan="1" colspan="1">Provide a different file name or path for the resulting .nsys-rep
                                    file. Default is the same name and path as the .qdstrm file
                                 </td>
                              </tr>
                           </table>
                        </div>
                     </div>
                  </div>
                  <div class="topic nested1" id="cli-analyze-mpi-codes"><a name="cli-analyze-mpi-codes" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#cli-analyze-mpi-codes" name="cli-analyze-mpi-codes" shape="rect">1.9.&nbsp;Using the CLI to Analyze MPI Codes</a></h3>
                     <div class="body">
                        <p class="p"></p>
                     </div>
                     <div class="topic nested2" id="cli-tracing-mpi-api-calls"><a name="cli-tracing-mpi-api-calls" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cli-tracing-mpi-api-calls" name="cli-tracing-mpi-api-calls" shape="rect">1.9.1.&nbsp;Tracing MPI API calls</a></h3>
                        <div class="body">
                           <p class="p">The <span class="ph">Nsight Systems</span> CLI has built-in API trace support for
                              Open MPI and MPICH based MPI implementations via <samp class="ph codeph">--trace=mpi</samp>.
                              It traces a subset of the MPI API, including blocking and non-blocking point-to-point and collective communication, and file
                              I/O operations.
                           </p>
                           <p class="p">If you require more control over the list of traced APIs or if you are
                              using a different MPI implementation, you can use the
                              <a class="xref" href="https://github.com/NVIDIA/cuda-profiler/tree/master/nvtx_pmpi_wrappers" target="_blank" shape="rect">NVTX wrappers for MPI</a> on GitHub.
                              Choose an NVTX domain name other than "MPI", since it is filtered out by
                              <span class="ph">Nsight Systems</span> when MPI tracing is not enabled.
                              Use the NVTX-instrumented MPI wrapper library as follows:
                           </p>
                           <p class="p"><samp class="ph codeph"><span class="ph">nsys</span> profile -e LD_PRELOAD=${PATH_TO_YOUR_NVTX_MPI_LIB} --trace=nvtx</samp></p>
                        </div>
                     </div>
                     <div class="topic nested2" id="cli-profiling-apps-launched-mpirun"><a name="cli-profiling-apps-launched-mpirun" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cli-profiling-apps-launched-mpirun" name="cli-profiling-apps-launched-mpirun" shape="rect">1.9.2.&nbsp;Using the CLI to Profile Applications Launched with mpirun</a></h3>
                        <div class="body">
                           <p class="p">This version of the <span class="ph">Nsight Systems</span> CLI supports concurrent use
                              of the <samp class="ph codeph"><span class="ph">nsys</span> profile</samp> command. Each instance
                              will create a separate report file.
                           </p>
                           <p class="p">You cannot use multiple instances of the interactive CLI concurrently, or use
                              the interactive CLI concurrently with <samp class="ph codeph"><span class="ph">nsys</span> profile
                                 </samp> in this version.
                           </p>
                           <p class="p"><span class="ph">Nsight Systems</span> can be used to profile applications launched with
                              <samp class="ph codeph">mpirun</samp> command. Since concurrent use of the CLI is supported
                              only when using the <samp class="ph codeph"><span class="ph">nsys</span> profile</samp> command,
                              <span class="ph">Nsight Systems</span> cannot profile each node from the GUI or from the
                              interactive CLI.
                           </p>
                           <p class="p">Profile all MPI ranks and put the data in one file (execution on single node only)</p><pre xml:space="preserve"><span class="ph">nsys</span> [<span class="ph">nsys</span> options] mpirun [mpirun options]</pre><p class="p">Profile all MPI ranks and put the data from each rank into a separate file:</p><pre xml:space="preserve">mpirun [mpirun options] <span class="ph">nsys</span> profile [<span class="ph">nsys</span> options]</pre><p class="p">To profile a single MPI process or a subset of MPI processes you can use a
                              wrapper script. The following script (called "wrap.sh") runs <span class="ph">nsys</span>
                              on rank 0 only.
                           </p><pre xml:space="preserve">#!/bin/bash

# Use $PMI_RANK for MPICH and $SLURM_PROCID with srun.
if [ $OMPI_COMM_WORLD_RANK -eq 0 ]; then
  <span class="ph">nsys</span> profile -e NSYS_MPI_STORE_TEAMS_PER_RANK=1 -t mpi "$@"
else
  "$@"
fi</pre><p class="p">Add appropriate profiling options to the script and execute it with
                              <samp class="ph codeph">mpirun [mpirun options] ./wrap.sh ./myapp [app options]</samp>.
                           </p>
                           <blockquote class="lq">
                              <div class="p">
                                 <div class="tablenoborder">
                                    <table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
                                       <tbody class="tbody">
                                          <tr class="row">
                                             <td class="entry nvidiagreen" align="center" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                                <p class="p">&nbsp;&nbsp;<strong class="ph b">Note:</strong>&nbsp;&nbsp;
                                                </p>
                                             </td>
                                             <td class="entry notebox" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                                <p class="p">If only a subset of MPI ranks is profiled, set the environment
                                                   variable <samp class="ph codeph">NSYS_MPI_STORE_TEAMS_PER_RANK=1</samp> to store
                                                   all members of custom MPI communicators per MPI rank. Otherwise,
                                                   the execution might hang or fail with an MPI error.
                                                </p>
                                             </td>
                                          </tr>
                                       </tbody>
                                    </table>
                                 </div>
                              </div>
                           </blockquote>
                           <p class="p"></p>
                           <p class="p hr"></p>
                           <p class="p"></p>
                        </div>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="gui-profiling"><a name="gui-profiling" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#gui-profiling" name="gui-profiling" shape="rect">2.&nbsp;Profiling from the GUI</a></h2>
                  <div class="body">
                     <p class="p"></p>
                  </div>
                  <div class="topic nested1" id="profiling-linux"><a name="profiling-linux" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#profiling-linux" name="profiling-linux" shape="rect">2.1.&nbsp;Profiling Linux Targets from the GUI</a></h3>
                     <div class="body">
                        <p class="p"></p>
                     </div>
                     <div class="topic nested2" id="linux-connect-target"><a name="linux-connect-target" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#linux-connect-target" name="linux-connect-target" shape="rect">2.1.1.&nbsp;Connecting to the Target Device</a></h3>
                        <div class="body">
                           <p class="p"><span class="ph">Nsight Systems</span> provides a simple interface to profile on
                              localhost or manage multiple connections to Linux or Windows based
                              devices via SSH. The network connections manager can be launched through
                              the device selection dropdown:
                           </p>
                           <p class="p">On x86_64:</p>
                           <blockquote class="lq">
                              <p class="p"><img class="image" src="graphics/device-list-empty-x86.png" height="-1" width="-1" alt="Empty device list"></img></p>
                           </blockquote>
                           <p class="p">On Tegra:</p>
                           <blockquote class="lq">
                              <p class="p"><img class="image" src="graphics/device-list-empty.png" alt="Empty device list"></img></p>
                           </blockquote>
                           <p class="p">The dialog has simple controls that allow adding, removing, and modifying
                              connections:
                           </p>
                           <blockquote class="lq">
                              <p class="p"><img class="image" src="graphics/network-connection.png" height="-1" width="-1" alt="Network connection"></img></p>
                           </blockquote>
                           <p class="p"><strong class="ph b">Security notice</strong>: SSH is only used to establish the initial
                              connection to a target device, perform checks, and upload necessary
                              files.  The actual profiling commands and data are transferred through a
                              raw, unencrypted socket. <span class="ph">Nsight Systems</span> should not be used
                              in a network setup where attacker-in-the-middle attack is possible, or
                              where untrusted parties may have network access to the target device.
                           </p>
                           <p class="p">While connecting to the target device, you will be prompted to input the
                              user's password. Please note that if you choose to remember the password,
                              it will be stored in plain text in the configuration file on the host.
                              Stored passwords are bound to the public key fingerprint of the remote
                              device.
                           </p>
                           <p class="p">The <strong class="ph b">No authentication</strong> option is useful for devices configured for
                              passwordless login using <samp class="ph codeph">root</samp> username. To enable such a
                              configuration, edit the file <samp class="ph codeph">/etc/ssh/sshd_config</samp> on the
                              target and specify the following option:
                           </p><pre xml:space="preserve">PermitRootLogin yes</pre><p class="p">Then set empty password using <samp class="ph codeph">passwd</samp> and restart the SSH
                              service with <samp class="ph codeph">service ssh restart</samp>.
                           </p>
                           <p class="p"><strong class="ph b">Open ports</strong>: The <span class="ph">Nsight Systems</span> daemon requires port
                              22 and port 45555 to be open for listening. You can confirm that these
                              ports are open with the following command:
                           </p><pre xml:space="preserve">sudo firewall-cmd --list-ports --permanent 
sudo firewall-cmd --reload</pre><p class="p">To open a port use the following command, skip <samp class="ph codeph">--permanent</samp>
                              option to open only for this session:
                           </p><pre xml:space="preserve">sudo firewall-cmd --permanent --add-port 45555/tcp 
sudo firewall-cmd --reload</pre><p class="p">Likewise, if you are running on a cloud system, you must open port 22 and
                              port 45555 for ingress.
                           </p>
                           <p class="p"><strong class="ph b">Kernel Version Number</strong> - To check for the version number of the
                              kernel support of <span class="ph">Nsight Systems</span> on a target device, run
                              the following command on the remote device:
                           </p><pre xml:space="preserve">cat /proc/quadd/version</pre><p class="p">Minimal supported version is <span class="ph">1.82</span>.
                           </p>
                           <p class="p">Additionally, presence of Netcat command (<samp class="ph codeph">nc</samp>) is
                              required on the target device.  For example, on Ubuntu this package can
                              be installed using the following command:
                           </p><pre xml:space="preserve">sudo apt-get install netcat-openbsd</pre></div>
                     </div>
                     <div class="topic nested2" id="linux-system-wide-profiling-options"><a name="linux-system-wide-profiling-options" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#linux-system-wide-profiling-options" name="linux-system-wide-profiling-options" shape="rect">2.1.2.&nbsp;System-Wide Profiling Options</a></h3>
                        <div class="body">
                           <p class="p"></p>
                        </div>
                        <div class="topic nested3" id="linux-x86-64"><a name="linux-x86-64" shape="rect">
                              <!-- --></a><h3 class="title topictitle2"><a href="#linux-x86-64" name="linux-x86-64" shape="rect">2.1.2.1.&nbsp;Linux x86_64</a></h3>
                           <div class="body">
                              <p class="p">System-wide profiling is available on x86 for Linux targets only when run with root
                                 privileges.
                              </p>
                              <p class="p"><strong class="ph b">Ftrace Events Collection</strong></p>
                              <p class="p">Select Ftrace events</p>
                              <blockquote class="lq">
                                 <p class="p"><img class="image" src="graphics/ftrace_trace_select.png" height="-1" width="-1" alt="Ftrace checkbox"></img></p>
                              </blockquote>
                              <p class="p">Choose which events you would like to collect.</p>
                              <blockquote class="lq">
                                 <p class="p"><img class="image" src="graphics/ftrace_options_gui.png" height="-1" width="-1" alt="System profiling options"></img></p>
                              </blockquote>
                              <p class="p"><strong class="ph b">GPU Context Switch Trace</strong></p>
                              <p class="p">Tracing of context switching on the GPU is enabled with driver
                                 r435.17 or higher.
                              </p>
                              <blockquote class="lq">
                                 <p class="p"><img class="image" src="graphics/select_gcs.png" height="-1" width="-1" alt="GCS checkbox"></img></p>
                              </blockquote>
                              <p class="p">Here is a screenshot showing three CUDA kernels running
                                 simultaneously in three different CUDA contexts on a single GPU.
                              </p>
                              <blockquote class="lq">
                                 <p class="p"><img class="image" src="graphics/gpu_context_switch.png" height="-1" width="-1" alt="GCS screenshot"></img></p>
                              </blockquote>
                           </div>
                        </div>
                        <div class="topic nested3" id="linux-for-tegra"><a name="linux-for-tegra" shape="rect">
                              <!-- --></a><h3 class="title topictitle2"><a href="#linux-for-tegra" name="linux-for-tegra" shape="rect">2.1.2.2.&nbsp;Linux for Tegra</a></h3>
                           <div class="body">
                              <blockquote class="lq">
                                 <p class="p"><img class="image" src="graphics/system-profiling-options.png" alt="System profiling options"></img></p>
                              </blockquote>
                              <p class="p"><strong class="ph b">Trace all processes</strong> – On compatible devices (with kernel module support
                                 version 1.107 or higher), this enables trace of all processes and threads in the system.
                                 Scheduler events from all tasks will be recorded.
                              </p>
                              <p class="p"><strong class="ph b">Collect PMU counters</strong> – This allows you to choose which PMU (Performance
                                 Monitoring Unit) counters <span class="ph">Nsight Systems</span> will sample. Enable specific counters
                                 when interested in correlating cache misses to functions in your application.
                              </p>
                           </div>
                        </div>
                     </div>
                     <div class="topic nested2" id="linux-target-sampling-options"><a name="linux-target-sampling-options" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#linux-target-sampling-options" name="linux-target-sampling-options" shape="rect">2.1.3.&nbsp;Target Sampling Options</a></h3>
                        <div class="body">
                           <p class="p">
                              Target sampling behavior is somewhat different for
                              <span class="ph">Nsight Systems Workstation Edition</span> and <span class="ph">Nsight Systems Embedded Platforms Edition</span>.
                           </p>
                        </div>
                        <div class="topic nested3" id="target-sampling-workstation"><a name="target-sampling-workstation" shape="rect">
                              <!-- --></a><h3 class="title topictitle2"><a href="#target-sampling-workstation" name="target-sampling-workstation" shape="rect">Target Sampling Options for Workstation</a></h3>
                           <div class="body">
                              <blockquote class="lq">
                                 <p class="p"><img class="image" src="graphics/target-sampling-options-x86.png" height="-1" width="-1" alt="Target sampling options"></img></p>
                              </blockquote>
                              <p class="p">Three different backtrace collections options are available when
                                 sampling CPU instruction pointers. Backtraces can be generated
                                 using Intel (c) Last Branch Record (LBR) registers. LBR backtraces
                                 generate minimal overhead but the backtraces have limited depth.
                                 Backtraces can also be generated using DWARF debug data. DWARF
                                 backtraces incur more overhead than LBR backtraces but have much
                                 better depth. Finally, backtraces can be generated using frame
                                 pointers. Frame pointer backtraces incur medium overhead and have
                                 good depth but only resolve frames in the portions of the
                                 application and its libraries (including 3rd party libraries) that
                                 were compiled with frame pointers enabled. Normally, frame pointers
                                 are disabled by default during compilation.
                              </p>
                              <p class="p">By default, <span class="ph">Nsight Systems</span> will use Intel(c) LBRs if
                                 available and fall back to using dwarf unwind if they are not.
                                 <strong class="ph b">Choose modes...</strong> will allow you to override the default.
                              </p>
                              <blockquote class="lq">
                                 <p class="p"><img class="image" src="graphics/backtrace-options.png" height="-1" width="-1" alt="Choose backtrace option"></img></p>
                              </blockquote>
                              <p class="p">
                                 The <strong class="ph b">Include child processes</strong> switch controls whether API
                                 tracing is only for the launched process, or for all existing and
                                 new child processes of the launched process. If you are running
                                 your application through a script, for example a bash script, you
                                 need to set this checkbox.
                              </p>
                              <p class="p">The <strong class="ph b">Include child processes</strong> switch does not control sampling
                                 in this version of <span class="ph">Nsight Systems</span>. The full process
                                 tree will be sampled regardless of this setting. This will be fixed
                                 in a future version of the product.
                              </p>
                              <p class="p"><span class="ph">Nsight Systems</span> can sample one process tree. Sampling
                                 here means interrupting each processor after a certain number of
                                 events and collecting an instruction pointer (IP)/backtrace sample
                                 if the processor is executing the profilee.
                              </p>
                              <p class="p">When sampling the CPU on a workstation target,
                                 <span class="ph">Nsight Systems</span> traces thread context switches and
                                 infers thread state as either Running or Blocked. Note that Blocked
                                 in the timeline indicates the thread may be Blocked (Interruptible)
                                 or Blocked (Uninterruptible). Blocked (Uninterruptible) often
                                 occurs when a thread has transitioned into the kernel and cannot be
                                 interrupted by a signal. Sampling can be enhanced with OS runtime
                                 libraries tracing; see <a class="xref" href="topics/os-runtime-trace" shape="rect">OS
                                    Runtime Libraries Trace</a> for more information.
                              </p>
                           </div>
                        </div>
                        <div class="topic nested3" id="target-sampling-tegra"><a name="target-sampling-tegra" shape="rect">
                              <!-- --></a><h3 class="title topictitle2"><a href="#target-sampling-tegra" name="target-sampling-tegra" shape="rect">Target Sampling Options for Embedded Linux</a></h3>
                           <div class="body">
                              <p class="p"><img class="image" src="graphics/target-sampling-options.png" alt="Target sampling options"></img></p>
                              <p class="p">Currently <span class="ph">Nsight Systems</span> can only sample one
                                 process. Sampling here means that the profilee will be stopped
                                 periodically, and backtraces of active threads will be recorded.
                              </p>
                              <p class="p">Most applications use stripped libraries. In this case, many symbols
                                 may stay unresolved. If unstripped libraries exist, paths to them
                                 can be specified using the <strong class="ph b">Symbol locations...</strong> button.
                                 Symbol resolution happens on host, and therefore does not affect
                                 performance of profiling on the target.
                              </p>
                              <p class="p">Additionally, debug versions of ELF files may be picked up from
                                 the target system. Refer to <a class="xref" href="topics/elf-debug" shape="rect">Debug
                                    Versions of ELF Files</a> for more information.
                              </p>
                           </div>
                        </div>
                     </div>
                     <div class="topic nested2" id="linux-hotkey-trace-start-stop"><a name="linux-hotkey-trace-start-stop" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#linux-hotkey-trace-start-stop" name="linux-hotkey-trace-start-stop" shape="rect">2.1.4.&nbsp;Hotkey Trace Start/Stop</a></h3>
                        <div class="body">
                           <p class="p"><span class="ph">Nsight Systems Workstation Edition</span> can use hotkeys to control profiling.
                              Press the hotkey to start and/or stop a trace session from within the
                              target application’s graphic window. This is useful when tracing games
                              and graphic applications that use fullscreen display. In these
                              scenarios switching to <span class="ph">Nsight Systems</span>' UI would
                              unnecessarily introduce the window manager's footprint into the trace.
                              To enable the use of Hotkey check the Hotkey checkbox in the project
                              settings page:
                           </p>
                           <blockquote class="lq">
                              <p class="p"><img class="image" src="graphics/hotkey-checkbox.png" alt="Hotkey checkbox"></img></p>
                           </blockquote>
                           <p class="p">The default hotkey is F12.</p>
                        </div>
                     </div>
                     <div class="topic nested2" id="linux-launch-processes"><a name="linux-launch-processes" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#linux-launch-processes" name="linux-launch-processes" shape="rect">2.1.5.&nbsp;Launching Processes</a></h3>
                        <div class="body">
                           <p class="p"><span class="ph">Nsight Systems</span> can launch new processes for profiling
                              on target devices. Profiler ensures that
                              all environment variables are set correctly to successfully collect
                              trace information
                           </p>
                           <blockquote class="lq">
                              <p class="p"><img class="image" src="graphics/process-launch.png" alt="Process: Launch"></img></p>
                           </blockquote>
                           <p class="p">The <strong class="ph b">Edit arguments...</strong> link will open an editor window, where
                              every command line argument is edited on a separate line. This is
                              convenient when arguments contain spaces or quotes.
                           </p>
                           <p class="p"></p>
                           <p class="p hr"></p>
                           <p class="p"></p>
                        </div>
                     </div>
                  </div>
                  <div class="topic nested1" id="profiling-windows"><a name="profiling-windows" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#profiling-windows" name="profiling-windows" shape="rect">2.2.&nbsp;Profiling Windows Targets from the GUI</a></h3>
                     <div class="body">
                        <p class="p">
                           Profiling on Windows devices is similar to the profiling on Linux devices.
                           Please refer to the <a class="xref" href="index.html#profiling-linux" shape="rect">Profiling Linux
                              Targets from the GUI</a> section for the detailed documentation and
                           connection information. The major differences on the platforms are listed below:
                        </p>
                     </div>
                     <div class="topic nested2" id="remote-windows"><a name="remote-windows" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#remote-windows" name="remote-windows" shape="rect">Remoting to a Windows Based Machine</a></h3>
                        <div class="body">
                           <p class="p">
                              To perform remote profiling to a target Windows based machines,
                              <a class="xref" href="https://docs.microsoft.com/en-us/windows-server/administration/openssh/openssh_install_firstuse" target="_blank" shape="rect">install and configure an OpenSSH Server on
                                 the target machine.</a></p>
                        </div>
                     </div>
                     <div class="topic nested2" id="hotkey-trace"><a name="hotkey-trace" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#hotkey-trace" name="hotkey-trace" shape="rect">Hotkey Trace Start/Stop</a></h3>
                        <div class="body">
                           <p class="p"><span class="ph">Nsight Systems Workstation Edition</span> can use hotkeys to control profiling.
                              Press the hotkey to start and/or stop a trace session from within the
                              target application’s graphic window. This is useful when tracing games
                              and graphic applications that use fullscreen display. In these
                              scenarios switching to <span class="ph">Nsight Systems</span>' UI would
                              unnecessarily introduce the window manager's footprint into the trace.
                              To enable the use of Hotkey check the Hotkey checkbox in the project
                              settings page:
                           </p>
                           <blockquote class="lq">
                              <p class="p"><img class="image" src="graphics/hotkey-checkbox.png" alt="Hotkey checkbox"></img></p>
                           </blockquote>
                           <p class="p">The default hotkey is F12.</p>
                           <p class="p"><strong class="ph b">Changing the Default Hotkey Binding</strong> - A different hotkey binding
                              can be configured by setting the HotKeyIntValue configuration field in the
                              config.ini file.
                           </p>
                           <p class="p">Set the decimal numeric identifier of the hotkey you would like to use for
                              triggering start/stop from the target app graphics window. The default
                              value is 123 which corresponds to 0x7B, or the F12 key.
                           </p>
                           <p class="p">Virtual key identifiers are detailed in
                              <a class="xref" href="https://docs.microsoft.com/en-us/windows/win32/inputdev/virtual-key-codes" target="_blank" shape="rect">MSDN's Virtual-Key Codes</a>.
                           </p>
                           <p class="p">Note that you must convert the hexadecimal values detailed in this page to
                              their decimal counterpart before using them in the file. For example, to
                              use the F1 key as a start/stop trace hotkey, use the following settings in
                              the config.ini file:
                           </p><pre xml:space="preserve">HotKeyIntValue=112</pre></div>
                     </div>
                     <div class="topic nested2" id="target-sampling-windows"><a name="target-sampling-windows" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#target-sampling-windows" name="target-sampling-windows" shape="rect">Target Sampling Options on Windows</a></h3>
                        <div class="body">
                           <blockquote class="lq">
                              <p class="p"><img class="image" src="graphics/target-sampling-options-windows.png" alt="Target sampling options"></img></p>
                           </blockquote>
                           <p class="p"><span class="ph">Nsight Systems</span> can sample one process tree. Sampling here
                              means interrupting each processor periodically. The sampling rate is
                              defined in the project settings and is either 100Hz, 1KHz (default value),
                              2Khz, 4KHz, or 8KHz.
                           </p>
                           <blockquote class="lq">
                              <p class="p"><img class="image" src="graphics/collect_thread_activity.png" alt="Thread activity option"></img></p>
                           </blockquote>
                           <p class="p">On Windows, <span class="ph">Nsight Systems</span> can collect thread activity of
                              one process tree. Collecting thread activity means that each thread
                              context switch event is logged and (optionally) a backtrace is collected
                              at the point that the thread is scheduled back for execution. Thread
                              states are displayed on the timeline.
                           </p>
                           <p class="p">If it was collected, the thread backtrace is displayed when hovering over
                              a region where the thread execution is blocked.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="windows-symbol-locations"><a name="windows-symbol-locations" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#windows-symbol-locations" name="windows-symbol-locations" shape="rect">Symbol Locations</a></h3>
                        <div class="body">
                           <p class="p">Symbol resolution happens on host, and therefore does not affect
                              performance of profiling on the target.
                           </p>
                           <p class="p">Press the <strong class="ph b">Symbol locations...</strong> button to open the <strong class="ph b">Configure
                                 debug symbols location</strong> dialog.
                           </p>
                           <blockquote class="lq">
                              <p class="p"><img class="image" src="graphics/symbols_location_dialog.png" alt="Configure debug symbols location"></img></p>
                           </blockquote>
                           <p class="p">Use this dialog to specify:</p>
                           <ul class="ul">
                              <li class="li">
                                 <p class="p">Paths of PDB files</p>
                              </li>
                              <li class="li">
                                 <p class="p">Symbols servers</p>
                              </li>
                              <li class="li">
                                 <p class="p">The location of the local symbol cache</p>
                              </li>
                           </ul>
                           <p class="p">To use a symbol server:</p>
                           <ol class="ol">
                              <li class="li">
                                 <p class="p">Install <strong class="ph b">Debugging Tools for Windows</strong>, a part of the <a class="xref" href="https://developer.microsoft.com/en-us/windows/downloads/windows-10-sdk" target="_blank" shape="rect">Windows 10 SDK</a>.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">Add the symbol server URL using the <strong class="ph b">Add Server</strong> button.
                                 </p>
                                 <p class="p">Information about Microsoft's public symbol server, which enables
                                    getting Windows operating system related debug symbols can be found
                                    <a class="xref" href="https://docs.microsoft.com/en-us/windows-hardware/drivers/debugger/microsoft-public-symbols" target="_blank" shape="rect">here</a>.
                                 </p>
                              </li>
                           </ol>
                           <p class="p"></p>
                           <p class="p hr"></p>
                           <p class="p"></p>
                        </div>
                     </div>
                  </div>
                  <div class="topic nested1" id="profiling-qnx"><a name="profiling-qnx" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#profiling-qnx" name="profiling-qnx" shape="rect">2.3.&nbsp;Profiling QNX Targets from the GUI</a></h3>
                     <div class="body">
                        <p class="p">Profiling on QNX devices is similar to the profiling on Linux devices.
                           Please refer to the <a class="xref" href="index.html#profiling-linux" shape="rect">Profiling Linux
                              Targets from the GUI</a> section for the detailed documentation.
                           The major differences on the platforms are listed below:
                        </p>
                        <ul class="ul">
                           <li class="li">
                              <p class="p">Backtrace sampling is not supported. Instead backtraces are collected for
                                 long OS runtime libraries calls. Please refer to the
                                 <a class="xref" href="topics/os-runtime-trace" shape="rect">OS Runtime Libraries Trace</a>
                                 section for the detailed documentation.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">CUDA support is limited to CUDA 9.0+</p>
                           </li>
                           <li class="li">
                              <p class="p">Filesystem on QNX device might be mounted read-only. In that case
                                 <span class="ph">Nsight Systems</span> is not able to install target-side binaries,
                                 required to run the profiling session. Please make sure that target
                                 filesystem is writable before connecting to QNX target. For example, make
                                 sure the following command works:
                              </p><pre xml:space="preserve">echo XX &gt; /xx &amp;&amp; ls -l /xx</pre></li>
                        </ul>
                        <p class="p"></p>
                        <p class="p hr"></p>
                        <p class="p"></p>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="exporter"><a name="exporter" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#exporter" name="exporter" shape="rect">3.&nbsp;Export Formats</a></h2>
                  <div class="body">
                     <p class="p"></p>
                  </div>
                  <div class="topic nested1" id="exporter-sqlite-schema"><a name="exporter-sqlite-schema" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#exporter-sqlite-schema" name="exporter-sqlite-schema" shape="rect">SQLite Schema Reference</a></h3>
                     <div class="body">
                        <p class="p"><span class="ph">Nsight Systems</span> has the ability to export SQLite database files
                           from the .nsys-rep results file. From the CLI,
                           use <samp class="ph codeph"><span class="ph">nsys</span> export</samp>. From the GUI, call
                           <samp class="ph codeph">File-&gt;Export...</samp>.
                        </p>
                        <p class="p"><strong class="ph b">Note:</strong> The .nsys-rep report format is the only data format for
                           <span class="ph">Nsight Systems</span> that should be considered forward compatible. The
                           SQLite schema can and will change in the future.
                        </p>
                        <p class="p">The schema for a concrete database can be obtained with the sqlite3 tool
                           built-in command <samp class="ph codeph">.schema</samp>. The sqlite3 tool can be located
                           in the Target or Host directory of your <span class="ph">Nsight Systems</span>
                           installation.
                        </p>
                        <p class="p"><strong class="ph b">Note:</strong> Currently tables are created lazily, and therefore not every
                           table described in the documentation will be present in a particular database.
                           This will change in a future version of the product. If you want a full schema
                           of all possible tables, use <samp class="ph codeph"><span class="ph">nsys</span> export --lazy=false
                              </samp> during export phase.
                        </p>
                        <p class="p">Currently, a table is created for each data type in the exported database.
                           Since usage patterns for exported data may vary greatly and no default use
                           cases have been established, no indexes or extra constraints are created.
                           Instead, refer to the SQLite Examples section for a list of common recipes.
                           This may change in a future version of the product.
                        </p>
                        <p class="p">To check the version of your exported SQLite file, check the value of
                           <samp class="ph codeph">EXPORT_SCHEMA_VERSION</samp> in the <samp class="ph codeph">EXPORT_META_DATA</samp>
                           table. The schema version is a common three-value major/minor/micro version
                           number. The first value, or major value, indicates the overall format of the
                           database, and is only changed if there is a major re-write or re-factor of
                           the entire database format. It is assumed that if the major version changes,
                           all scripts or queries will break. The middle, or minor, version is changed
                           anytime there is a more localized, but potentially breaking change, such as
                           renaming an existing column, or changing the type of an existing column. The
                           last, or micro version is changed any time there are additions, such as a
                           new table or column, that should not introduce any breaking change when used
                           with well-written, best-practices queries.
                        </p>
                        <p class="p">This is the schema as of the 2021.5 release, schema version 2.7.1.</p><pre xml:space="preserve">
CREATE TABLE StringIds (
    -- Consolidation of repetitive string values.

    id                          INTEGER   NOT NULL   PRIMARY KEY,      -- ID reference value.
    value                       TEXT      NOT NULL                     -- String value.
);
CREATE TABLE ThreadNames (
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the thread name.
    priority                    INTEGER,                               -- Priority of the thread.
    globalTid                   INTEGER                                -- Serialized GlobalId.
);
CREATE TABLE ProcessStreams (
    globalPid                   INTEGER   NOT NULL,                    -- Serialized GlobalId.
    filenameId                  INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the file name.
    contentId                   INTEGER   NOT NULL   REFERENCES StringIds(id) -- StringId of the stream content.
);
CREATE TABLE TARGET_INFO_SYSTEM_ENV (
    globalVid                   INTEGER   NOT NULL,                    -- Serialized GlobalId.
    devStateName                TEXT      NOT NULL,                    -- Device state name.
    name                        TEXT      NOT NULL,                    -- Property name.
    nameEnum                    INTEGER   NOT NULL,                    -- Property enum value.
    value                       TEXT      NOT NULL                     -- Property value.
);
CREATE TABLE TARGET_INFO_SESSION_START_TIME (
    utcEpochNs                  INTEGER,                               -- UTC Epoch timestamp at start of the capture (ns).
    utcTime                     TEXT,                                  -- Start of the capture in UTC.
    localTime                   TEXT                                   -- Start of the capture in local time of target.
);
CREATE TABLE ANALYSIS_DETAILS (
    -- Details about the analysis session.

    globalVid                   INTEGER   NOT NULL,                    -- Serialized GlobalId.
    duration                    INTEGER   NOT NULL,                    -- The total time span of the entire trace (ns).
    startTime                   INTEGER   NOT NULL,                    -- Trace start timestamp in nanoseconds.
    stopTime                    INTEGER   NOT NULL                     -- Trace stop timestamp in nanoseconds.
);
CREATE TABLE TARGET_INFO_GPU (
    vmId                        INTEGER   NOT NULL,                    -- Serialized GlobalId.
    id                          INTEGER   NOT NULL,                    -- Device ID.
    name                        TEXT,                                  -- Device name.
    busLocation                 TEXT,                                  -- PCI bus location.
    isDiscrete                  INTEGER,                               -- True if discrete, false if integrated.
    l2CacheSize                 INTEGER,                               -- Size of L2 cache (B).
    totalMemory                 INTEGER,                               -- Total amount of memory on the device (B).
    memoryBandwidth             INTEGER,                               -- Amount of memory transferred (B).
    clockRate                   INTEGER,                               -- Clock frequency (Hz).
    smCount                     INTEGER,                               -- Number of multiprocessors on the device.
    pwGpuId                     INTEGER,                               -- PerfWorks GPU ID.
    uuid                        TEXT,                                  -- Device UUID.
    luid                        INTEGER,                               -- Device LUID.
    chipName                    TEXT,                                  -- Chip name.
    cuDevice                    INTEGER,                               -- CUDA device ID.
    ctxswDevPath                TEXT,                                  -- GPU context switch device node path.
    ctrlDevPath                 TEXT,                                  -- GPU control device node path.
    revision                    INTEGER,                               -- Revision number.
    nodeMask                    INTEGER,                               -- Device node mask.
    constantMemory              INTEGER,                               -- Memory available on device for __constant__ variables (B).
    maxIPC                      INTEGER,                               -- Maximum instructions per count.
    maxRegistersPerBlock        INTEGER,                               -- Maximum number of 32-bit registers available per block.
    maxShmemPerBlock            INTEGER,                               -- Maximum optin shared memory per block.
    maxShmemPerBlockOptin       INTEGER,                               -- Maximum optin shared memory per block.
    maxShmemPerSm               INTEGER,                               -- Maximum shared memory available per multiprocessor (B).
    maxRegistersPerSm           INTEGER,                               -- Maximum number of 32-bit registers available per multiprocessor.
    threadsPerWarp              INTEGER,                               -- Warp size in threads.
    asyncEngines                INTEGER,                               -- Number of asynchronous engines.
    maxWarpsPerSm               INTEGER,                               -- Maximum number of warps per multiprocessor.
    maxBlocksPerSm              INTEGER,                               -- Maximum number of blocks per multiprocessor.
    maxThreadsPerBlock          INTEGER,                               -- Maximum number of threads per block.
    maxBlockDimX                INTEGER,                               -- Maximum X-dimension of a block.
    maxBlockDimY                INTEGER,                               -- Maximum Y-dimension of a block.
    maxBlockDimZ                INTEGER,                               -- Maximum Z-dimension of a block.
    maxGridDimX                 INTEGER,                               -- Maximum X-dimension of a grid.
    maxGridDimY                 INTEGER,                               -- Maximum Y-dimension of a grid.
    maxGridDimZ                 INTEGER,                               -- Maximum Z-dimension of a grid.
    computeMajor                INTEGER,                               -- Major compute capability version number.
    computeMinor                INTEGER,                               -- Minor compute capability version number.
    smMajor                     INTEGER,                               -- Major multiprocessor version number.
    smMinor                     INTEGER                                -- Minor multiprocessor version number.
);
CREATE TABLE TARGET_INFO_XMC_SPEC (
    vmId                        INTEGER   NOT NULL,                    -- Serialized GlobalId.
    clientId                    INTEGER   NOT NULL,                    -- Client ID.
    type                        TEXT      NOT NULL,                    -- Client type.
    name                        TEXT      NOT NULL,                    -- Client name.
    groupId                     TEXT      NOT NULL                     -- Client group ID.
);
CREATE TABLE TARGET_INFO_PROCESS (
    processId                   INTEGER   NOT NULL,                    -- Process ID.
    openGlVersion               TEXT      NOT NULL,                    -- OpenGL version.
    correlationId               INTEGER   NOT NULL,                    -- Correlation ID of the kernel.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id) -- StringId of the function name.
);
CREATE TABLE TARGET_INFO_NVTX_CUDA_DEVICE (
    name                        TEXT      NOT NULL,                    -- CUDA device name assigned using NVTX.
    hwId                        INTEGER   NOT NULL,                    -- Hardware ID.
    vmId                        INTEGER   NOT NULL,                    -- VM ID.
    deviceId                    INTEGER   NOT NULL                     -- Device ID.
);
CREATE TABLE TARGET_INFO_NVTX_CUDA_CONTEXT (
    name                        TEXT      NOT NULL,                    -- CUDA context name assigned using NVTX.
    hwId                        INTEGER   NOT NULL,                    -- Hardware ID.
    vmId                        INTEGER   NOT NULL,                    -- VM ID.
    processId                   INTEGER   NOT NULL,                    -- Process ID.
    deviceId                    INTEGER   NOT NULL,                    -- Device ID.
    contextId                   INTEGER   NOT NULL                     -- Context ID.
);
CREATE TABLE TARGET_INFO_NVTX_CUDA_STREAM (
    name                        TEXT      NOT NULL,                    -- CUDA stream name assigned using NVTX.
    hwId                        INTEGER   NOT NULL,                    -- Hardware ID.
    vmId                        INTEGER   NOT NULL,                    -- VM ID.
    processId                   INTEGER   NOT NULL,                    -- Process ID.
    deviceId                    INTEGER   NOT NULL,                    -- Device ID.
    contextId                   INTEGER   NOT NULL,                    -- Context ID.
    streamId                    INTEGER   NOT NULL                     -- Stream ID.
);
CREATE TABLE TARGET_INFO_CUDA_NULL_STREAM (
    streamId                    INTEGER   NOT NULL,                    -- Stream ID.
    hwId                        INTEGER   NOT NULL,                    -- Hardware ID.
    vmId                        INTEGER   NOT NULL,                    -- VM ID.
    processId                   INTEGER   NOT NULL,                    -- Process ID.
    deviceId                    INTEGER   NOT NULL,                    -- Device ID.
    contextId                   INTEGER   NOT NULL                     -- Context ID.
);
CREATE TABLE TARGET_INFO_CUDA_STREAM (
    streamId                    INTEGER   NOT NULL,                    -- Stream ID.
    hwId                        INTEGER   NOT NULL,                    -- Hardware ID.
    vmId                        INTEGER   NOT NULL,                    -- VM ID.
    processId                   INTEGER   NOT NULL,                    -- Process ID.
    contextId                   INTEGER   NOT NULL,                    -- Context ID.
    priority                    INTEGER   NOT NULL,                    -- Priority of the stream.
    flag                        INTEGER   NOT NULL                     -- Flags for stream creation.
);
CREATE TABLE TARGET_INFO_WDDM_CONTEXTS (
    context                     INTEGER   NOT NULL,
    engineType                  INTEGER   NOT NULL,
    nodeOrdinal                 INTEGER   NOT NULL,
    friendlyName                TEXT      NOT NULL
);
CREATE TABLE EXPORT_META_DATA (
    -- information about nsys export process

    name                        TEXT      NOT NULL,                    -- Name of export meta-data record
    value                       TEXT                                   -- Value of export meta-data record
);
CREATE TABLE CUPTI_ACTIVITY_KIND_MEMCPY (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    deviceId                    INTEGER   NOT NULL,                    -- Device ID.
    contextId                   INTEGER   NOT NULL,                    -- Context ID.
    streamId                    INTEGER   NOT NULL,                    -- Stream ID.
    correlationId               INTEGER   REFERENCES CUPTI_ACTIVITY_KIND_RUNTIME(correlationId), -- ID used to identify the API call that launched the event.
    globalPid                   INTEGER,                               -- Serialized GlobalId.
    bytes                       INTEGER   NOT NULL,                    -- Number of bytes transferred (B).
    copyKind                    INTEGER   NOT NULL,                    -- Type of the memory copy. See enum CUDA_MEMCPY_KIND.
    deprecatedSrcId             INTEGER,                               -- Deprecated, use srcDeviceId instead.
    srcKind                     INTEGER,                               -- Source memory kind See enum CUDA_MEMOPR_MEMORY_KIND.
    dstKind                     INTEGER,                               -- Destination memory kind. See enum CUDA_MEMOPR_MEMORY_KIND.
    srcDeviceId                 INTEGER,                               -- Source device ID.
    srcContextId                INTEGER,                               -- Source context ID.
    dstDeviceId                 INTEGER,                               -- Destination device ID.
    dstContextId                INTEGER,                               -- Destination context ID.
    migrationCause              INTEGER,                               -- UVM migration cause ID. See CUpti_ActivityUnifiedMemoryMigrationCause.
    graphNodeId                 INTEGER   REFERENCES CUDA_GRAPH_EVENTS(graphNodeId), -- CUDA graph node creation ID.
    virtualAddress              INTEGER                                -- Virtual base address of the page/s being transferred.
);
CREATE TABLE CUPTI_ACTIVITY_KIND_MEMSET (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    deviceId                    INTEGER   NOT NULL,                    -- Device ID.
    contextId                   INTEGER   NOT NULL,                    -- Context ID.
    streamId                    INTEGER   NOT NULL,                    -- Stream ID.
    correlationId               INTEGER   REFERENCES CUPTI_ACTIVITY_KIND_RUNTIME(correlationId), -- ID used to identify the API call that launched the event.
    globalPid                   INTEGER,                               -- Serialized GlobalId.
    value                       INTEGER   NOT NULL,                    -- Value assigned to memory.
    bytes                       INTEGER   NOT NULL,                    -- Number of bytes set (B).
    graphNodeId                 INTEGER   REFERENCES CUDA_GRAPH_EVENTS(graphNodeId), -- CUDA graph node creation ID.
    memKind                     INTEGER                                -- Type of memory being set. See enum CUDA_MEMOPR_MEMORY_KIND.
);
CREATE TABLE CUPTI_ACTIVITY_KIND_KERNEL (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    deviceId                    INTEGER   NOT NULL,                    -- Device ID.
    contextId                   INTEGER   NOT NULL,                    -- Context ID.
    streamId                    INTEGER   NOT NULL,                    -- Stream ID.
    correlationId               INTEGER   REFERENCES CUPTI_ACTIVITY_KIND_RUNTIME(correlationId), -- ID used to identify the API call that launched the event.
    globalPid                   INTEGER,                               -- Serialized GlobalId.
    demangledName               INTEGER   NOT NULL,                    -- StringId of the kernel name
    shortName                   INTEGER   NOT NULL,                    -- StringId of the short function name without arguments.
    mangledName                 INTEGER,                               -- StringId of the mangled function name.
    launchType                  INTEGER,                               -- Launch type of the kernel. See docs for specifics.
    cacheConfig                 INTEGER,                               -- Cache configuration used for the kernel. See enum CUfunc_cache.
    registersPerThread          INTEGER   NOT NULL,                    -- Number of registers required for each thread executing the kernel.
    gridX                       INTEGER   NOT NULL,                    -- X-dimension grid size.
    gridY                       INTEGER   NOT NULL,                    -- Y-dimension grid size.
    gridZ                       INTEGER   NOT NULL,                    -- Z-dimension grid size.
    blockX                      INTEGER   NOT NULL,                    -- X-dimension block size.
    blockY                      INTEGER   NOT NULL,                    -- Y-dimension block size.
    blockZ                      INTEGER   NOT NULL,                    -- Z-dimension block size.
    staticSharedMemory          INTEGER   NOT NULL,                    -- Static shared memory allocated for the kernel (B).
    dynamicSharedMemory         INTEGER   NOT NULL,                    -- Dynamic shared memory reserved for the kernel (B).
    localMemoryPerThread        INTEGER   NOT NULL,                    -- Amount of local memory reserved for each thread (B).
    localMemoryTotal            INTEGER   NOT NULL,                    -- Total amount of local memory reserved for the kernel (B).
    gridId                      INTEGER   NOT NULL,                    -- Unique grid ID of the kernel assigned at runtime.
    sharedMemoryExecuted        INTEGER,                               -- Shared memory size set by the driver.
    graphNodeId                 INTEGER   REFERENCES CUDA_GRAPH_EVENTS(graphNodeId), -- CUDA graph node creation ID.
    sharedMemoryLimitConfig     INTEGER                                -- Shared memory limit per block config. See enum CUpti_FuncShmemLimitConfig.
);
CREATE TABLE CUPTI_ACTIVITY_KIND_CUDA_EVENT (
    deviceId                    INTEGER   NOT NULL,                    -- Device ID.
    contextId                   INTEGER   NOT NULL,                    -- Context ID.
    streamId                    INTEGER   NOT NULL,                    -- Stream ID.
    correlationId               INTEGER,                               -- Correlation ID of the event record API to which this result is associated.
    globalPid                   INTEGER,                               -- Serialized GlobalId.
    eventId                     INTEGER   NOT NULL                     -- Event ID for which the event record API is called.
);
CREATE TABLE CUPTI_ACTIVITY_KIND_SYNCHRONIZATION (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    deviceId                    INTEGER   NOT NULL,                    -- Device ID.
    contextId                   INTEGER   NOT NULL,                    -- Context ID.
    streamId                    INTEGER   NOT NULL,                    -- Stream ID.
    correlationId               INTEGER,                               -- Correlation ID of the API to which this result is associated.
    globalPid                   INTEGER,                               -- Serialized GlobalId.
    syncType                    INTEGER   NOT NULL,                    -- Type of synchronization. See enum CUpti_ActivitySynchronizationType.
    eventId                     INTEGER   NOT NULL                     -- Event ID for which the synchronization API is called.
);
CREATE TABLE CUPTI_ACTIVITY_KIND_RUNTIME (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- CUDA event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- ID used to identify events that this function call has triggered.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the function name.
    returnValue                 INTEGER   NOT NULL,                    -- Return value of the function call.
    callchainId                 INTEGER   REFERENCES CUDA_CALLCHAINS(id) -- ID of the attached callchain.
);
CREATE TABLE CUDNN_EVENTS (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- CUDA event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id) -- StringId of the function name.
);
CREATE TABLE CUBLAS_EVENTS (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- CUDA event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id) -- StringId of the function name.
);
CREATE TABLE CUDA_GRAPH_EVENTS (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- CUDA event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the function name.
    graphNodeId                 INTEGER   NOT NULL   REFERENCES CUDA_GRAPH_EVENTS(graphNodeId), -- CUDA graph node creation ID.
    originalGraphNodeId         INTEGER                                -- Reference to the original graph node ID, if cloned node.
);
CREATE TABLE CUDA_UM_CPU_PAGE_FAULT_EVENTS (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    globalPid                   INTEGER   NOT NULL,                    -- Serialized GlobalId.
    address                     INTEGER   NOT NULL,                    -- Virtual address of the page that faulted.
    originalFaultPc             INTEGER,                               -- Program counter of the CPU instruction that caused the page fault.
    CpuInstruction              INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the function name.
    module                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the module name.
    unresolvedFaultPc           INTEGER                                -- True if the program counter was not resolved.
);
CREATE TABLE CUDA_UM_GPU_PAGE_FAULT_EVENTS (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalPid                   INTEGER   NOT NULL,                    -- Serialized GlobalId.
    deviceId                    INTEGER   NOT NULL,                    -- Device ID.
    address                     INTEGER   NOT NULL,                    -- Virtual address of the page that faulted.
    numberOfPageFaults          INTEGER   NOT NULL,                    -- Number of page faults for the same page.
    faultAccessType             INTEGER   NOT NULL                     -- Type of Unified memory accessed. See enum CUpti_ActivityUnifiedMemoryAccessType.
);
CREATE TABLE CUDA_GPU_MEMORY_USAGE_EVENTS (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    globalPid                   INTEGER   NOT NULL,                    -- Serialized GlobalId.
    deviceId                    INTEGER   NOT NULL,                    -- Device ID.
    contextId                   INTEGER   NOT NULL,                    -- Context ID.
    address                     INTEGER   NOT NULL,                    -- Virtual address of the allocation/deallocation.
    pc                          INTEGER   NOT NULL,                    -- Program counter of the allocation/deallocation.
    bytes                       INTEGER   NOT NULL,                    -- Number of bytes allocated/deallocated (B).
    memKind                     INTEGER   NOT NULL,                    -- Type of memory being allocated/deallocated. See enum CUDA_MEMOPR_MEMORY_KIND.
    memoryOperationType         INTEGER   NOT NULL,                    -- 0 if memory was allocated and 1 if memory was deallocated.
    name                        TEXT,                                  -- Variable name, if available.
    correlationId               INTEGER                                -- ID used to identify the API call that launched the event.
);
CREATE TABLE CUDA_CALLCHAINS (
    id                          INTEGER   NOT NULL,                    -- Part of PRIMARY KEY (id, stackDepth).
    symbol                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the function name.
    module                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the module name.
    unresolved                  INTEGER,                               -- True if the symbol was not resolved.
    originalIP                  INTEGER,                               -- Instruction pointer value.
    stackDepth                  INTEGER   NOT NULL,                    -- Zero-base index of the given function in call stack.

    PRIMARY KEY (id, stackDepth)
);
CREATE TABLE NVTX_EVENTS (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER,                               -- Event end timestamp (ns).
    eventType                   INTEGER   NOT NULL,                    -- NVTX event type enum value. See docs for specifics.
    rangeId                     INTEGER,                               -- Correlation ID returned from a nvtxRangeStart call.
    category                    INTEGER,                               -- User-controlled ID that can be used to group events.
    color                       INTEGER,                               -- Encoded ARGB color value.
    text                        TEXT,                                  -- Optional text message for non registered strings.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    endGlobalTid                INTEGER,                               -- Serialized GlobalId. See docs for specifics.
    textId                      INTEGER   REFERENCES StringIds(id),    -- StringId of the NVTX domain registered string.
    domainId                    INTEGER,                               -- User-controlled ID that can be used to group events.
    uint64Value                 INTEGER,                               -- One of possible payload value union members.
    int64Value                  INTEGER,                               -- One of possible payload value union members.
    doubleValue                 REAL,                                  -- One of possible payload value union members.
    uint32Value                 INTEGER,                               -- One of possible payload value union members.
    int32Value                  INTEGER,                               -- One of possible payload value union members.
    floatValue                  REAL,                                  -- One of possible payload value union members.
    jsonTextId                  INTEGER,                               -- One of possible payload value union members.
    jsonText                    TEXT                                   -- One of possible payload value union members.
);
CREATE TABLE OPENGL_API (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenGL event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    endGlobalTid                INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- First ID matching an API call to GPU workloads.
    endCorrelationId            INTEGER,                               -- Last ID matching an API call to GPU workloads.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the first function name.
    endNameId                   INTEGER   REFERENCES StringIds(id),    -- StringId of the last function name.
    returnValue                 INTEGER   NOT NULL,                    -- Return value of the function call.
    frameId                     INTEGER,                               -- Index of the graphics frame starting from 1.
    contextId                   INTEGER,                               -- Context ID.
    gpu                         INTEGER,                               -- GPU index.
    display                     INTEGER                                -- Display ID.
);
CREATE TABLE OPENGL_WORKLOAD (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenGL event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    endGlobalTid                INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- First ID matching an API call to GPU workloads.
    endCorrelationId            INTEGER,                               -- Last ID matching an API call to GPU workloads.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the first function name.
    endNameId                   INTEGER   REFERENCES StringIds(id),    -- StringId of the last function name.
    returnValue                 INTEGER   NOT NULL,                    -- Return value of the function call.
    frameId                     INTEGER,                               -- Index of the graphics frame starting from 1.
    contextId                   INTEGER,                               -- Context ID.
    gpu                         INTEGER,                               -- GPU index.
    display                     INTEGER                                -- Display ID.
);
CREATE TABLE KHR_DEBUG_EVENTS (
    eventClass                  INTEGER   NOT NULL,                    -- KHR event class enum value. See docs for specifics.
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER,                               -- Event end timestamp (ns).
    textId                      INTEGER   REFERENCES StringIds(id),    -- StringId of the debug marker/group text string.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    source                      INTEGER,                               -- KHR source enum value. See docs for specifics.
    khrdType                    INTEGER,                               -- KHR type enum value. See docs for specifics.
    id                          INTEGER,                               -- KHR event ID.
    severity                    INTEGER,                               -- KHR severity enum value. See docs for specifics.
    correlationId               INTEGER,                               -- ID used to correlate KHR CPU trace to GPU trace.
    context                     INTEGER                                -- Context ID.
);
CREATE TABLE OSRT_API (
    -- OS runtime libraries traced to gather information about low-level userspace APIs.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OSRT event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the function name.
    returnValue                 INTEGER   NOT NULL,                    -- Return value of the function call.
    nestingLevel                INTEGER,                               -- Zero-base index of the nesting level.
    callchainId                 INTEGER   NOT NULL   REFERENCES OSRT_CALLCHAINS(id) -- ID of the attached callchain.
);
CREATE TABLE OSRT_CALLCHAINS (
    -- Callchains attached to OSRT events, depending on selected profiling settings.

    id                          INTEGER   NOT NULL,                    -- Part of PRIMARY KEY (id, stackDepth).
    symbol                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the function name.
    module                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the module name.
    kernelMode                  INTEGER,                               -- True if kernel mode.
    thumbCode                   INTEGER,                               -- True if thumb code.
    unresolved                  INTEGER,                               -- True if the symbol was not resolved.
    specialEntry                INTEGER,                               -- True if artifical entry added during processing callchain.
    originalIP                  INTEGER,                               -- Instruction pointer value.
    unwindMethod                INTEGER   REFERENCES UnwindMethodType(number), -- Instruction pointer value.
    stackDepth                  INTEGER   NOT NULL,                    -- Zero-base index of the given function in call stack.

    PRIMARY KEY (id, stackDepth)
);
CREATE TABLE UnwindMethodType (
    number                      INTEGER   PRIMARY KEY,
    name                        TEXT      NOT NULL
);
CREATE TABLE PROFILER_OVERHEAD (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the function name.
    returnValue                 INTEGER   NOT NULL                     -- Return value of the function call.
);
CREATE TABLE SCHED_EVENTS (
    -- Thread scheduling events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    cpu                         INTEGER   NOT NULL,                    -- ID of CPU this thread was scheduled in or out.
    isSchedIn                   INTEGER   NOT NULL,                    -- 0 if thread was scheduled out, non-zero otherwise.
    globalTid                   INTEGER                                -- Serialized GlobalId.
);
CREATE TABLE COMPOSITE_EVENTS (
    -- Thread sampling events.

    id                          INTEGER   NOT NULL   PRIMARY KEY,      -- ID of the composite event.
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    cpu                         INTEGER,                               -- ID of CPU this thread was running on.
    threadState                 INTEGER,                               -- Thread state at the moment of sampling. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    cpuCycles                   INTEGER   NOT NULL                     -- Value of Performance Monitoring Unit (PMU) counter.
);
CREATE TABLE SAMPLING_CALLCHAINS (
    -- Callchain entries obtained from composite events, used to construct function table views.

    id                          INTEGER   NOT NULL   REFERENCES COMPOSITE_EVENTS(id), -- ID of the attached composite event.
    symbol                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the function name.
    module                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the module name.
    kernelMode                  INTEGER,                               -- True if kernel mode.
    thumbCode                   INTEGER,                               -- True if thumb code.
    unresolved                  INTEGER,                               -- True if the symbol was not resolved.
    specialEntry                INTEGER,                               -- True if artifical entry added during processing callchain.
    originalIP                  INTEGER,                               -- Instruction pointer value.
    unwindMethod                INTEGER   REFERENCES UnwindMethodType(number), -- Instruction pointer value.
    stackDepth                  INTEGER   NOT NULL,                    -- Zero-base index of the given function in call stack.

    PRIMARY KEY (id, stackDepth)
);
CREATE TABLE SLI_QUERIES (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- SLI event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    frameId                     INTEGER   NOT NULL,                    -- Index of the graphics frame starting from 1.
    occQueryIssued              INTEGER   NOT NULL,                    -- Occlusion query issued.
    occQueryAsked               INTEGER   NOT NULL,                    -- Occlusion query asked.
    eventQueryIssued            INTEGER   NOT NULL,                    -- Event query issued.
    eventQueryAsked             INTEGER   NOT NULL,                    -- Event query asked.
    numberOfTransferEvents      INTEGER   NOT NULL,                    -- Number of transfer events.
    amountOfTransferredData     INTEGER   NOT NULL                     -- Cumulative size of resource data that was transferred.
);
CREATE TABLE SLI_P2P (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- SLI event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    frameId                     INTEGER   NOT NULL,                    -- Index of the graphics frame starting from 1.
    transferSkipped             INTEGER   NOT NULL,                    -- Number of transfers that were skipped.
    srcGpu                      INTEGER   NOT NULL,                    -- Source GPU ID.
    dstGpu                      INTEGER   NOT NULL,                    -- Destination GPU ID.
    numSubResources             INTEGER   NOT NULL,                    -- Number of sub-resources to transfer.
    resourceSize                INTEGER   NOT NULL,                    -- Size of resource.
    subResourceIdx              INTEGER   NOT NULL,                    -- Sub-resource index.
    smplWidth                   INTEGER,                               -- Sub-resource surface width in samples.
    smplHeight                  INTEGER,                               -- Sub-resource surface height in samples.
    smplDepth                   INTEGER,                               -- Sub-resource surface depth in samples.
    bytesPerElement             INTEGER,                               -- Number of bytes per element.
    dxgiFormat                  INTEGER,                               -- Resource data format. See DXGI_FORMAT enum.
    logSurfaceNames             TEXT,                                  -- Surface name.
    transferInfo                INTEGER,                               -- SLI transfer info enum value.  See docs for specifics.
    isEarlyPushManagedByNvApi   INTEGER,                               -- True if early push managed by NVAPI. False otherwise.
    useAsyncP2pForResolve       INTEGER,                               -- True if async Peer-to-Peer used for resolve. False otherwise.
    transferFuncName            TEXT,                                  -- "A - BE" for asynchronous transfer, "S - BE" for synchronous transfer.
    regimeName                  TEXT,                                  -- Name of the regime scope that includes the resource.
    debugName                   TEXT,                                  -- Debug name assigned to the resource by the application code.
    bindType                    TEXT                                   -- Bind type.
);
CREATE TABLE SLI_STATS (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- SLI event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    countComplexFrames          INTEGER   NOT NULL,                    -- Complex frames count.
    countStats                  INTEGER   NOT NULL,                    -- Number of frame statistics collected for the inactive-time histogram.
    totalInactiveTime           INTEGER   NOT NULL,                    -- Total inactive time (μs).
    minPbSize                   INTEGER   NOT NULL,                    -- Min push buffer size.
    maxPbSize                   INTEGER   NOT NULL,                    -- Max push buffer size.
    totalPbSize                 INTEGER   NOT NULL                     -- Total push buffer size.
);
CREATE TABLE DX12_API (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- DX12 event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- First ID matching an API call to GPU workloads.
    endCorrelationId            INTEGER,                               -- Last ID matching an API call to GPU workloads.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the function name.
    shortContextId              INTEGER,                               -- Short form of the COM interface object address.
    frameId                     INTEGER,                               -- Index of the graphics frame starting from 1.
    color                       INTEGER,                               -- Encoded ARGB color value.
    textId                      INTEGER,                               -- StringId of the PIX marker text string.
    commandListType             INTEGER,                               -- Type of command list. See enum D3D12_COMMAND_LIST_TYPE.
    objectName                  TEXT,                                  -- StringId of the D3D12 object name.
    longContextId               INTEGER                                -- Long form of the COM interface object address.
);
CREATE TABLE DX12_WORKLOAD (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- DX12 event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- First ID matching an API call to GPU workloads.
    endCorrelationId            INTEGER,                               -- Last ID matching an API call to GPU workloads.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the function name.
    shortContextId              INTEGER,                               -- Short form of the COM interface object address.
    frameId                     INTEGER,                               -- Index of the graphics frame starting from 1.
    gpu                         INTEGER,                               -- GPU index.
    color                       INTEGER,                               -- Encoded ARGB color value.
    textId                      INTEGER,                               -- StringId of the PIX marker text string.
    commandListType             INTEGER,                               -- Type of command list. See enum D3D12_COMMAND_LIST_TYPE.
    objectName                  TEXT,                                  -- StringId of the D3D12 object name.
    longContextId               INTEGER                                -- Long form of the COM interface object address.
);
CREATE TABLE VULKAN_API (
    id                          INTEGER   NOT NULL   PRIMARY KEY,
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- Vulkan event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- First ID matching an API call to GPU workloads.
    endCorrelationId            INTEGER,                               -- Last ID matching an API call to GPU workloads.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the function name.
    contextId                   INTEGER                                -- Short form of the interface object address.
);
CREATE TABLE VULKAN_WORKLOAD (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- Vulkan event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- First ID matching an API call to GPU workloads.
    endCorrelationId            INTEGER,                               -- Last ID matching an API call to GPU workloads.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the function name.
    gpu                         INTEGER,                               -- GPU index.
    contextId                   INTEGER,                               -- Short form of the interface object address.
    color                       INTEGER,                               -- Encoded ARGB color value.
    textId                      INTEGER                                -- StringId of the VULKAN CPU debug marker text string.
);
CREATE TABLE VULKAN_DEBUG_API (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- Vulkan event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- First ID matching an API call to GPU workloads.
    endCorrelationId            INTEGER,                               -- Last ID matching an API call to GPU workloads.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the function name.
    contextId                   INTEGER,                               -- Short form of the interface object address.
    color                       INTEGER,                               -- Encoded ARGB color value.
    textId                      INTEGER                                -- StringId of the VULKAN CPU debug marker text string.
);
CREATE TABLE VULKAN_PIPELINE_CREATION_EVENTS (
    id                          INTEGER   NOT NULL   PRIMARY KEY,      -- ID of the pipeline creation event.
    duration                    INTEGER,                               -- Event duration (ns).
    flags                       INTEGER,
    traceEventId                INTEGER   NOT NULL   REFERENCES VULKAN_API(id) -- ID of the attached vulkan API.
);
CREATE TABLE VULKAN_PIPELINE_STAGE_EVENTS (
    id                          INTEGER   NOT NULL   PRIMARY KEY,      -- ID of the pipeline stage event.
    duration                    INTEGER,                               -- Event duration (ns).
    flags                       INTEGER,                               -- Vulkan flag enum value. See docs for specifics.
    creationEventId             INTEGER   NOT NULL   REFERENCES VULKAN_PIPELINE_CREATION_EVENTS(id) -- ID of the attached pipeline creation event.
);
CREATE TABLE FECS_EVENTS (
    tag                         INTEGER   NOT NULL,                    -- Context switch event type. See docs for specifics.
    vmId                        INTEGER   NOT NULL,                    -- VM ID.
    seqNo                       INTEGER   NOT NULL,                    -- Sequential event number.
    contextId                   INTEGER   NOT NULL,                    -- Context ID.
    timestamp                   INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    globalPid                   INTEGER,                               -- Serialized GlobalId.
    gpuId                       INTEGER                                -- GPU index.
);
CREATE TABLE ETW_EVENTS_DEPRECATED_TABLE (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    providerName                INTEGER   REFERENCES StringIds(id),    -- StringId of tracing events provider name.
    taskName                    INTEGER   REFERENCES StringIds(id),    -- StringId of the event task name.
    description                 INTEGER   REFERENCES StringIds(id)     -- StringId of the decoded event value.
);
CREATE TABLE ETW_PROVIDERS (
    -- Names and identifiers of ETW providers captured in the report.

    providerId                  INTEGER   NOT NULL   PRIMARY KEY,      -- Provider ID.
    providerNameId              INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of provider name.
    guid                        TEXT      NOT NULL                     -- ETW Provider GUID.
);
CREATE TABLE ETW_TASKS (
    -- Names and identifiers of ETW tasks captured in the report.

    taskNameId                  INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of task name.
    taskId                      INTEGER   NOT NULL,                    -- The event task ID.
    providerId                  INTEGER   NOT NULL                     -- Provider ID.
);
CREATE TABLE ETW_EVENTS (
    -- Raw ETW events captured in the report.

    timestamp                   INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    processId                   INTEGER,                               -- Process ID.
    threadId                    INTEGER,                               -- Thread ID.
    providerId                  INTEGER   NOT NULL,                    -- Provider ID.
    taskId                      INTEGER   NOT NULL,                    -- The event task ID.
    eventId                     INTEGER   NOT NULL,                    -- Event ID.
    version                     INTEGER   NOT NULL,                    -- The event version.
    opcode                      INTEGER,                               -- The event opcode.
    data                        TEXT      NOT NULL                     -- JSON encoded event data.
);
CREATE TABLE OPENMP_EVENT_KIND_THREAD (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    threadId                    INTEGER,                               -- Internal thread sequence starting from 1.
    threadType                  INTEGER                                -- See enum ompt_thread_t.
);
CREATE TABLE OPENMP_EVENT_KIND_PARALLEL (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    parallelId                  INTEGER,                               -- Internal parallel region sequence starting from 1.
    parentTaskId                INTEGER                                -- ID for task that creates this parallel region.
);
CREATE TABLE OPENMP_EVENT_KIND_SYNC_REGION_WAIT (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    parallelId                  INTEGER,                               -- ID of the parallel region that this event belongs to.
    taskId                      INTEGER,                               -- ID of the task that this event belongs to.
    kind                        INTEGER                                -- See enum ompt_sync_region_t.
);
CREATE TABLE OPENMP_EVENT_KIND_SYNC_REGION (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    parallelId                  INTEGER,                               -- ID of the parallel region that this event belongs to.
    taskId                      INTEGER,                               -- ID of the task that this event belongs to.
    kind                        INTEGER                                -- See enum ompt_sync_region_t.
);
CREATE TABLE OPENMP_EVENT_KIND_TASK (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    parallelId                  INTEGER,                               -- ID of the parallel region that this event belongs to.
    taskId                      INTEGER,                               -- ID of the task that this event belongs to.
    kind                        INTEGER                                -- See enum ompt_task_flag_t.
);
CREATE TABLE OPENMP_EVENT_KIND_MASTER (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    parallelId                  INTEGER,                               -- ID of the parallel region that this event belongs to.
    taskId                      INTEGER                                -- ID of the task that this event belongs to.
);
CREATE TABLE OPENMP_EVENT_KIND_REDUCTION (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    parallelId                  INTEGER,                               -- ID of the parallel region that this event belongs to.
    taskId                      INTEGER                                -- ID of the task that this event belongs to.
);
CREATE TABLE OPENMP_EVENT_KIND_TASK_CREATE (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    parentTaskId                INTEGER,                               -- ID of the parent task that is creating a new task.
    newTaskId                   INTEGER                                -- ID of the new task that is being created.
);
CREATE TABLE OPENMP_EVENT_KIND_TASK_SCHEDULE (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    parallelId                  INTEGER,                               -- ID of the parallel region that this event belongs to.
    priorTaskId                 INTEGER,                               -- ID of the task that is being switched out.
    priorTaskStatus             INTEGER,                               -- See enum ompt_task_status_t.
    nextTaskId                  INTEGER                                -- ID of the task that is being switched in.
);
CREATE TABLE OPENMP_EVENT_KIND_CANCEL (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    taskId                      INTEGER                                -- ID of the task that is being cancelled.
);
CREATE TABLE OPENMP_EVENT_KIND_MUTEX_WAIT (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    kind                        INTEGER,                               -- See enum ompt_mutex_t.
    waitId                      INTEGER,                               -- ID indicating the object being waited.
    taskId                      INTEGER                                -- ID of the task that this event belongs to.
);
CREATE TABLE OPENMP_EVENT_KIND_CRITICAL_SECTION (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    kind                        INTEGER,                               -- See enum ompt_mutex_t.
    waitId                      INTEGER                                -- ID indicating the object being held.
);
CREATE TABLE OPENMP_EVENT_KIND_MUTEX_RELEASED (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    kind                        INTEGER,                               -- See enum ompt_mutex_t.
    waitId                      INTEGER,                               -- ID indicating the object being released.
    taskId                      INTEGER                                -- ID of the task that this event belongs to.
);
CREATE TABLE OPENMP_EVENT_KIND_LOCK_INIT (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    kind                        INTEGER,                               -- See enum ompt_mutex_t.
    waitId                      INTEGER                                -- ID indicating object being created/destroyed.
);
CREATE TABLE OPENMP_EVENT_KIND_LOCK_DESTROY (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    kind                        INTEGER,                               -- See enum ompt_mutex_t.
    waitId                      INTEGER                                -- ID indicating object being created/destroyed.
);
CREATE TABLE OPENMP_EVENT_KIND_WORKSHARE (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    kind                        INTEGER,                               -- See enum ompt_work_t.
    parallelId                  INTEGER,                               -- ID of the parallel region that this event belongs to.
    taskId                      INTEGER,                               -- ID of the task that this event belongs to.
    count                       INTEGER                                -- Measure of the quantity of work involved in the region.
);
CREATE TABLE OPENMP_EVENT_KIND_DISPATCH (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    kind                        INTEGER,                               -- See enum ompt_dispatch_t
    parallelId                  INTEGER,                               -- ID of the parallel region that this event belongs to.
    taskId                      INTEGER                                -- ID of the task that this event belongs to.
);
CREATE TABLE OPENMP_EVENT_KIND_FLUSH (
    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- OpenMP event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- Currently unused.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    eventKind                   INTEGER,                               -- OpenMP event kind enum value. See docs for specifics.
    threadId                    INTEGER                                -- ID of the thread that this event belongs to.
);
CREATE TABLE D3D11_PIX_DEBUG_API (
    -- D3D11 debug marker events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- First ID matching an API call to GPU workloads.
    endCorrelationId            INTEGER,                               -- Last ID matching an API call to GPU workloads.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the function name.
    shortContextId              INTEGER,                               -- Short form of the COM interface object address.
    frameId                     INTEGER,                               -- Index of the graphics frame starting from 1.
    color                       INTEGER,                               -- Encoded ARGB color value.
    textId                      INTEGER                                -- StringId of the PIX marker text string.
);
CREATE TABLE D3D12_PIX_DEBUG_API (
    -- D3D12 debug marker events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    correlationId               INTEGER,                               -- First ID matching an API call to GPU workloads.
    endCorrelationId            INTEGER,                               -- Last ID matching an API call to GPU workloads.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the function name.
    shortContextId              INTEGER,                               -- Short form of the COM interface object address.
    frameId                     INTEGER,                               -- Index of the graphics frame starting from 1.
    color                       INTEGER,                               -- Encoded ARGB color value.
    textId                      INTEGER,                               -- StringId of the PIX marker text string.
    commandListType             INTEGER,                               -- Type of command list. See enum D3D12_COMMAND_LIST_TYPE.
    objectName                  TEXT,                                  -- StringId of the D3D12 object name.
    longContextId               INTEGER                                -- Long form of the COM interface object address.
);
CREATE TABLE WDDM_EVICT_ALLOCATION_EVENTS (
    -- Raw ETW EvictAllocation events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    allocationHandle            INTEGER   NOT NULL                     -- Global allocation handle.
);
CREATE TABLE WDDM_PAGING_QUEUE_PACKET_START_EVENTS (
    -- Raw ETW PagingQueuePacketStart events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    dxgDevice                   INTEGER,                               -- Address of an IDXGIDevice.
    dxgAdapter                  INTEGER,                               -- Address of an IDXGIAdapter.
    pagingQueue                 INTEGER   NOT NULL,                    -- Address of the paging queue.
    pagingQueuePacket           INTEGER   NOT NULL,                    -- Address of the paging queue packet.
    sequenceId                  INTEGER   NOT NULL,                    -- Internal sequence starting from 0.
    alloc                       INTEGER,                               -- Allocation handle.
    vidMmOpType                 INTEGER   NOT NULL,                    -- VIDMM operation type enum value. See docs for specifics.
    pagingQueueType             INTEGER   NOT NULL                     -- Paging queue type enum value. See docs for specifics.
);
CREATE TABLE WDDM_PAGING_QUEUE_PACKET_STOP_EVENTS (
    -- Raw ETW PagingQueuePacketStop events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    pagingQueue                 INTEGER   NOT NULL,                    -- Address of the paging queue.
    pagingQueuePacket           INTEGER   NOT NULL,                    -- Address of the paging queue packet.
    sequenceId                  INTEGER   NOT NULL                     -- Internal sequence starting from 0.
);
CREATE TABLE WDDM_PAGING_QUEUE_PACKET_INFO_EVENTS (
    -- Raw ETW PagingQueuePacketInfo events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    pagingQueue                 INTEGER   NOT NULL,                    -- Address of the paging queue.
    pagingQueuePacket           INTEGER   NOT NULL,                    -- Address of the paging queue packet.
    sequenceId                  INTEGER   NOT NULL                     -- Internal sequence starting from 0.
);
CREATE TABLE WDDM_QUEUE_PACKET_START_EVENTS (
    -- Raw ETW QueuePacketStart events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    context                     INTEGER   NOT NULL,                    -- The context ID of WDDM queue.
    dmaBufferSize               INTEGER   NOT NULL,                    -- The dma buffer size.
    dmaBuffer                   INTEGER   NOT NULL,                    -- The reported address of dma buffer.
    queuePacket                 INTEGER   NOT NULL,                    -- The address of queue packet.
    progressFenceValue          INTEGER   NOT NULL,                    -- The fence value.
    packetType                  INTEGER   NOT NULL,                    -- The type of the packet.
    submitSequence              INTEGER   NOT NULL,                    -- Internal sequence starting from 1.
    allocationListSize          INTEGER   NOT NULL,                    -- The number of allocations referenced.
    patchLocationListSize       INTEGER   NOT NULL,                    -- The number of patch locations.
    present                     INTEGER   NOT NULL,                    -- True or False if the packet is a present packet.
    engineType                  INTEGER   NOT NULL,                    -- The engine type.
    syncObject                  INTEGER                                -- The address of fence object.
);
CREATE TABLE WDDM_QUEUE_PACKET_STOP_EVENTS (
    -- Raw ETW QueuePacketStop events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    context                     INTEGER   NOT NULL,                    -- The context ID of WDDM queue.
    queuePacket                 INTEGER   NOT NULL,                    -- The address of queue packet.
    packetType                  INTEGER   NOT NULL,                    -- The type of the packet.
    submitSequence              INTEGER   NOT NULL,                    -- Internal sequence starting from 1.
    preempted                   INTEGER   NOT NULL,                    -- True or False if the packet is preempted.
    timeouted                   INTEGER   NOT NULL,                    -- True or False if the packet is timeouted.
    engineType                  INTEGER   NOT NULL                     -- The engine type.
);
CREATE TABLE WDDM_QUEUE_PACKET_INFO_EVENTS (
    -- Raw ETW QueuePacketInfo events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    context                     INTEGER   NOT NULL,                    -- The context ID of WDDM queue.
    packetType                  INTEGER   NOT NULL,                    -- The address of queue packet.
    submitSequence              INTEGER   NOT NULL,                    -- Internal sequence starting from 1.
    engineType                  INTEGER   NOT NULL                     -- The engine type.
);
CREATE TABLE WDDM_DMA_PACKET_START_EVENTS (
    -- Raw ETW DmaPacketStart events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    context                     INTEGER   NOT NULL,                    -- The context ID of WDDM queue.
    queuePacketContext          INTEGER   NOT NULL,                    -- The queue packet context.
    uliSubmissionId             INTEGER   NOT NULL,                    -- The queue packet submission ID.
    dmaBuffer                   INTEGER   NOT NULL,                    -- The reported address of dma buffer.
    packetType                  INTEGER   NOT NULL,                    -- The type of the packet.
    ulQueueSubmitSequence       INTEGER   NOT NULL,                    -- Internal sequence starting from 1.
    quantumStatus               INTEGER   NOT NULL,                    -- The quantum Status.
    engineType                  INTEGER   NOT NULL                     -- The engine type.
);
CREATE TABLE WDDM_DMA_PACKET_STOP_EVENTS (
    -- Raw ETW DmaPacketStop events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    context                     INTEGER   NOT NULL,                    -- The context ID of WDDM queue.
    uliCompletionId             INTEGER   NOT NULL,                    -- The queue packet completion ID.
    packetType                  INTEGER   NOT NULL,                    -- The type of the packet.
    ulQueueSubmitSequence       INTEGER   NOT NULL,                    -- Internal sequence starting from 1.
    preempted                   INTEGER   NOT NULL,                    -- True or False if the packet is preempted.
    engineType                  INTEGER   NOT NULL                     -- The engine type.
);
CREATE TABLE WDDM_DMA_PACKET_INFO_EVENTS (
    -- Raw ETW DmaPacketInfo events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    context                     INTEGER   NOT NULL,                    -- The context ID of WDDM queue.
    uliCompletionId             INTEGER   NOT NULL,                    -- The queue packet completion ID.
    faultedVirtualAddress       INTEGER   NOT NULL,                    -- The virtual address of faulted process.
    faultedProcessHandle        INTEGER   NOT NULL,                    -- The address of faulted process.
    packetType                  INTEGER   NOT NULL,                    -- The type of the packet.
    ulQueueSubmitSequence       INTEGER   NOT NULL,                    -- Internal sequence starting from 1.
    interruptType               INTEGER   NOT NULL,                    -- The DMA interrupt type.
    quantumStatus               INTEGER   NOT NULL,                    -- The quantum Status.
    pageFaultFlags              INTEGER   NOT NULL,                    -- The page fault flag ID.
    engineType                  INTEGER   NOT NULL                     -- The engine type.
);
CREATE TABLE WDDM_HW_QUEUE_EVENTS (
    -- Raw ETW HwQueueStart events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    context                     INTEGER   NOT NULL,                    -- The context ID of WDDM queue.
    hwQueue                     INTEGER   NOT NULL,                    -- The address of HW queue.
    parentDxgHwQueue            INTEGER   NOT NULL                     -- The address of parent Dxg HW queue.
);
CREATE TABLE NVVIDEO_ENCODER_API (
    -- NV Video Encoder API traced to gather information about NVIDIA Video Codek SDK Encoder APIs.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- NVIDIA Video event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id) -- StringId of the function name.
);
CREATE TABLE NVVIDEO_DECODER_API (
    -- NV Video Encoder API traced to gather information about NVIDIA Video Codek SDK Decoder APIs.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    eventClass                  INTEGER   NOT NULL,                    -- NVIDIA Video event class enum value. See docs for specifics.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id) -- StringId of the function name.
);
CREATE TABLE GPU_MEMORY_BUDGET_EVENTS (
    -- Raw ETW VidMmProcessBudgetChange events.

    timestamp                   INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    newBudget                   INTEGER,                               -- The new budget size in bytes.
    segmentGroup                INTEGER                                -- The segment group ID.
);
CREATE TABLE GPU_MEMORY_USAGE_EVENTS (
    -- Raw ETW VidMmProcessUsageChange events.

    timestamp                   INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    newUsage                    INTEGER,                               -- The new usage size in bytes.
    segmentGroup                INTEGER                                -- The segment group ID.
);
CREATE TABLE DEMOTED_BYTES_EVENTS (
    -- Raw ETW VidMmProcessDemotedCommitmentChange events.

    timestamp                   INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    commitment                  INTEGER                                -- Total demoted bytes.
);
CREATE TABLE TOTAL_BYTES_RESIDENT_IN_SEGMENT_EVENTS (
    -- Raw ETW TotalBytesResidentInSegment events.

    timestamp                   INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    totalBytesResident          INTEGER,                               -- Total bytes resident in segment.
    segmentGroup                INTEGER,                               -- The segment group ID.
    segmentId                   INTEGER                                -- The segment ID.
);
CREATE TABLE NV_LOAD_BALANCE_MASTER_EVENTS (
    -- Raw ETW NV-wgf2um LoadBalanceMaster events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER   NOT NULL,                    -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    eventId                     INTEGER   NOT NULL,                    -- Event ID.
    task                        TEXT      NOT NULL,                    -- The task name.
    frameCount                  INTEGER   NOT NULL,                    -- The frame ID.
    frameTime                   REAL      NOT NULL,                    -- Frame duration.
    averageFrameTime            REAL      NOT NULL,                    -- Average of frame duration.
    averageLatency              REAL      NOT NULL,                    -- Average of latency.
    minLatency                  REAL      NOT NULL,                    -- The minimum latency.
    averageQueuedFrames         REAL      NOT NULL,                    -- Average number of queued frames.
    totalActiveMs               REAL      NOT NULL,                    -- Total active time in milliseconds.
    totalIdleMs                 REAL      NOT NULL,                    -- Total idle time in milliseconds.
    idlePercent                 REAL      NOT NULL,                    -- The percentage of idle time.
    isGPUAlmostOneFrameAhead    INTEGER   NOT NULL                     -- True or False if GPU is almost one frame ahead.
);
CREATE TABLE NV_LOAD_BALANCE_EVENTS (
    -- Raw ETW NV-wgf2um LoadBalance events.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    globalTid                   INTEGER   NOT NULL,                    -- Serialized GlobalId.
    gpu                         INTEGER   NOT NULL,                    -- GPU index.
    eventId                     INTEGER   NOT NULL,                    -- Event ID.
    task                        TEXT      NOT NULL,                    -- The task name.
    averageFPS                  REAL      NOT NULL,                    -- Average frame per second.
    queuedFrames                REAL      NOT NULL,                    -- The amount of queued frames.
    averageQueuedFrames         REAL      NOT NULL,                    -- Average number of queued frames.
    currentCPUTime              REAL      NOT NULL,                    -- The current CPU time.
    averageCPUTime              REAL      NOT NULL,                    -- Average CPU time.
    averageStallTime            REAL      NOT NULL,                    -- Average of stall time.
    averageCPUIdleTime          REAL      NOT NULL,                    -- Average CPU idle time.
    isGPUAlmostOneFrameAhead    INTEGER   NOT NULL                     -- True or False if GPU is almost one frame ahead.
);
CREATE TABLE PROCESSES (
    -- Names and identifiers of processes captured in the report.

    globalPid                   INTEGER,                               -- Serialized GlobalId.
    pid                         INTEGER,                               -- The process ID.
    name                        TEXT                                   -- The process name.
);
CREATE TABLE CUPTI_ACTIVITY_KIND_OPENACC_DATA (
    -- OpenACC data events collected using CUPTI.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    eventKind                   INTEGER   NOT NULL,                    -- CUPTI OpenACC event kind. See enum CUpti_OpenAccEventKind.
    DeviceType                  INTEGER   NOT NULL,                    -- Device type. See enum acc_device_t.
    lineNo                      INTEGER   NOT NULL,                    -- Line number of the directive or program construct.
    cuDeviceId                  INTEGER   NOT NULL,                    -- CUDA device ID. Valid only if deviceType is acc_device_nvidia.
    cuContextId                 INTEGER   NOT NULL,                    -- CUDA context ID. Valid only if deviceType is acc_device_nvidia.
    cuStreamId                  INTEGER   NOT NULL,                    -- CUDA stream ID. Valid only if deviceType is acc_device_nvidia.
    srcFile                     INTEGER   REFERENCES StringIds(id),    -- StringId of the source file name or path.
    funcName                    INTEGER   REFERENCES StringIds(id),    -- StringId of the function in which the event occurred.
    bytes                       INTEGER,                               -- Number of bytes.
    varName                     INTEGER   REFERENCES StringIds(id)     -- StringId of the variable name.
);
CREATE TABLE CUPTI_ACTIVITY_KIND_OPENACC_LAUNCH (
    -- OpenACC launch events collected using CUPTI.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    eventKind                   INTEGER   NOT NULL,                    -- CUPTI OpenACC event kind. See enum CUpti_OpenAccEventKind.
    DeviceType                  INTEGER   NOT NULL,                    -- Device type. See enum acc_device_t.
    lineNo                      INTEGER   NOT NULL,                    -- Line number of the directive or program construct.
    cuDeviceId                  INTEGER   NOT NULL,                    -- CUDA device ID. Valid only if deviceType is acc_device_nvidia.
    cuContextId                 INTEGER   NOT NULL,                    -- CUDA context ID. Valid only if deviceType is acc_device_nvidia.
    cuStreamId                  INTEGER   NOT NULL,                    -- CUDA stream ID. Valid only if deviceType is acc_device_nvidia.
    srcFile                     INTEGER   REFERENCES StringIds(id),    -- StringId of the source file name or path.
    funcName                    INTEGER   REFERENCES StringIds(id),    -- StringId of the function in which the event occurred.
    numGangs                    INTEGER,                               -- Number of gangs created for this kernel launch.
    numWorkers                  INTEGER,                               -- Number of workers created for this kernel launch.
    vectorLength                INTEGER,                               -- Number of vector lanes created for this kernel launch.
    kernelName                  INTEGER   REFERENCES StringIds(id)     -- StringId of the kernel name.
);
CREATE TABLE CUPTI_ACTIVITY_KIND_OPENACC_OTHER (
    -- OpenACC other events collected using CUPTI.

    start                       INTEGER   NOT NULL,                    -- Event start timestamp (ns).
    end                         INTEGER   NOT NULL,                    -- Event end timestamp (ns).
    nameId                      INTEGER   NOT NULL   REFERENCES StringIds(id), -- StringId of the event name.
    globalTid                   INTEGER,                               -- Serialized GlobalId.
    eventKind                   INTEGER   NOT NULL,                    -- CUPTI OpenACC event kind. See enum CUpti_OpenAccEventKind.
    DeviceType                  INTEGER   NOT NULL,                    -- Device type. See enum acc_device_t.
    lineNo                      INTEGER   NOT NULL,                    -- Line number of the directive or program construct.
    cuDeviceId                  INTEGER   NOT NULL,                    -- CUDA device ID. Valid only if deviceType is acc_device_nvidia.
    cuContextId                 INTEGER   NOT NULL,                    -- CUDA context ID. Valid only if deviceType is acc_device_nvidia.
    cuStreamId                  INTEGER   NOT NULL,                    -- CUDA stream ID. Valid only if deviceType is acc_device_nvidia.
    srcFile                     INTEGER   REFERENCES StringIds(id),    -- StringId of the source file name or path.
    funcName                    INTEGER   REFERENCES StringIds(id)     -- StringId of the function in which the event occurred.
);
</pre></div>
                  </div>
                  <div class="topic nested1" id="exporter-sqlite-event-values"><a name="exporter-sqlite-event-values" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#exporter-sqlite-event-values" name="exporter-sqlite-event-values" shape="rect">SQLite Schema Event Values</a></h3>
                     <div class="body">
                        <p class="p">Here are the set values stored in enums in the <span class="ph">Nsight Systems</span>
                           SQLite schema
                        </p>
                        <p class="p"><strong class="ph b">CUDA Event Class Values</strong></p><pre xml:space="preserve">
0 - TRACE_PROCESS_EVENT_CUDA_RUNTIME
1 - TRACE_PROCESS_EVENT_CUDA_DRIVER
13 - TRACE_PROCESS_EVENT_CUDA_EGL_DRIVER
28 - TRACE_PROCESS_EVENT_CUDNN
29 - TRACE_PROCESS_EVENT_CUBLAS
33 - TRACE_PROCESS_EVENT_CUDNN_START
34 - TRACE_PROCESS_EVENT_CUDNN_FINISH
35 - TRACE_PROCESS_EVENT_CUBLAS_START
36 - TRACE_PROCESS_EVENT_CUBLAS_FINISH
67 - TRACE_PROCESS_EVENT_CUDABACKTRACE
77 - TRACE_PROCESS_EVENT_CUDA_GRAPH_NODE_CREATION    
    </pre><p class="p">See <a class="xref" href="https://docs.nvidia.com/cupti/Cupti/index.html" target="_blank" shape="rect">CUPTI documentation</a> for detailed information on
                           collected event and data types.
                        </p>
                        <p class="p"><strong class="ph b">NVTX Event Type Values</strong></p><pre xml:space="preserve">
33 - NvtxCategory
34 - NvtxMark
39 - NvtxThread
59 - NvtxPushPopRange
60 - NvtxStartEndRange
75 - NvtxDomainCreate
76 - NvtxDomainDestroy
    </pre><p class="p">The difference between text and textId columns is that if an NVTX event
                           message was passed via call to nvtxDomainRegisterString function, then the
                           message will be available through textId field, otherwise the text field will
                           contain the message if it was provided.
                        </p>
                        <p class="p"><strong class="ph b">OpenGL Events</strong></p>
                        <p class="p">KHR event class values</p><pre xml:space="preserve">
62 - KhrDebugPushPopRange
63 - KhrDebugGpuPushPopRange
    </pre><p class="p">KHR source kind values</p><pre xml:space="preserve">
0x8249 - GL_DEBUG_SOURCE_THIRD_PARTY
0x824A - GL_DEBUG_SOURCE_APPLICATION
    </pre><p class="p">KHR type values</p><pre xml:space="preserve">
0x824C - GL_DEBUG_TYPE_ERROR
0x824D - GL_DEBUG_TYPE_DEPRECATED_BEHAVIOR
0x824E - GL_DEBUG_TYPE_UNDEFINED_BEHAVIOR
0x824F - GL_DEBUG_TYPE_PORTABILITY
0x8250 - GL_DEBUG_TYPE_PERFORMANCE
0x8251 - GL_DEBUG_TYPE_OTHER
0x8268 - GL_DEBUG_TYPE_MARKER
0x8269 - GL_DEBUG_TYPE_PUSH_GROUP
0x826A - GL_DEBUG_TYPE_POP_GROUP
   </pre><p class="p">KHR severity values</p><pre xml:space="preserve">
0x826B - GL_DEBUG_SEVERITY_NOTIFICATION
0x9146 - GL_DEBUG_SEVERITY_HIGH
0x9147 - GL_DEBUG_SEVERITY_MEDIUM
0x9148 - GL_DEBUG_SEVERITY_LOW
   </pre><p class="p"><strong class="ph b">OSRT Event Class Values</strong></p>
                        <p class="p">OS runtime libraries can be traced to gather information about low-level
                           userspace APIs. This traces the system call wrappers and thread
                           synchronization interfaces exposed by the C runtime and POSIX Threads (pthread)
                           libraries. This does not perform a complete runtime library API trace, but
                           instead focuses on the functions that can take a long time to execute, or
                           could potentially cause your thread be unscheduled from the CPU while waiting
                           for an event to complete.
                        </p>
                        <p class="p">OSRT events may have callchains attached to them, depending on selected
                           profiling settings. In such cases, one can use callchainId column to select
                           relevant callchains from OSRT_CALLCHAINS table
                        </p>
                        <p class="p">OSRT event class values</p><pre xml:space="preserve">
27 - TRACE_PROCESS_EVENT_OS_RUNTIME
31 - TRACE_PROCESS_EVENT_OS_RUNTIME_START
32 - TRACE_PROCESS_EVENT_OS_RUNTIME_FINISH
   </pre><p class="p"><strong class="ph b">DX12 Event Class Values</strong></p><pre xml:space="preserve">
41 - TRACE_PROCESS_EVENT_DX12_API
42 - TRACE_PROCESS_EVENT_DX12_WORKLOAD
43 - TRACE_PROCESS_EVENT_DX12_START
44 - TRACE_PROCESS_EVENT_DX12_FINISH
52 - TRACE_PROCESS_EVENT_DX12_DISPLAY
59 - TRACE_PROCESS_EVENT_DX12_CREATE_OBJECT
   </pre><p class="p"><strong class="ph b">PIX Event Class Values</strong></p><pre xml:space="preserve">
65 - TRACE_PROCESS_EVENT_DX12_DEBUG_API
75 - TRACE_PROCESS_EVENT_DX11_DEBUG_API
   </pre><p class="p"><strong class="ph b">Vulkan Event Class Values</strong></p><pre xml:space="preserve">
53 - TRACE_PROCESS_EVENT_VULKAN_API
54 - TRACE_PROCESS_EVENT_VULKAN_WORKLOAD
55 - TRACE_PROCESS_EVENT_VULKAN_START
56 - TRACE_PROCESS_EVENT_VULKAN_FINISH
60 - TRACE_PROCESS_EVENT_VULKAN_CREATE_OBJECT
66 - TRACE_PROCESS_EVENT_VULKAN_DEBUG_API
   </pre><p class="p"><strong class="ph b">Vulkan Flags</strong></p><pre xml:space="preserve">
VALID_BIT = 0x00000001
CACHE_HIT_BIT = 0x00000002
BASE_PIPELINE_ACCELERATION_BIT = 0x00000004
   </pre><p class="p"><strong class="ph b">SLI Event Class Values</strong></p><pre xml:space="preserve">
62 - TRACE_PROCESS_EVENT_SLI
63 - TRACE_PROCESS_EVENT_SLI_START
64 - TRACE_PROCESS_EVENT_SLI_FINISH
   </pre><p class="p"><strong class="ph b">SLI Transfer Info Values</strong></p><pre xml:space="preserve">
0 - P2P_SKIPPED
1 - P2P_EARLY_PUSH
2 - P2P_PUSH_FAILED
3 - P2P_2WAY_OR_PULL
4 - P2P_PRESENT
5 - P2P_DX12_INIT_PUSH_ON_WRITE
    </pre><p class="p"><strong class="ph b">WDDM Event Values</strong></p>
                        <p class="p">VIDMM operation type values</p><pre xml:space="preserve">
0 - None
101 - RestoreSegments
102 - PurgeSegments
103 - CleanupPrimary
104 - AllocatePagingBufferResources
105 - FreePagingBufferResources
106 - ReportVidMmState
107 - RunApertureCoherencyTest
108 - RunUnmapToDummyPageTest
109 - DeferredCommand
110 - SuspendMemorySegmentAccess
111 - ResumeMemorySegmentAccess
112 - EvictAndFlush
113 - CommitVirtualAddressRange
114 - UncommitVirtualAddressRange
115 - DestroyVirtualAddressAllocator
116 - PageInDevice
117 - MapContextAllocation
118 - InitPagingProcessVaSpace
200 - CloseAllocation
202 - ComplexLock
203 - PinAllocation
204 - FlushPendingGpuAccess
205 - UnpinAllocation
206 - MakeResident
207 - Evict
208 - LockInAperture
209 - InitContextAllocation
210 - ReclaimAllocation
211 - DiscardAllocation
212 - SetAllocationPriority
1000 - EvictSystemMemoryOfferList
   </pre><p class="p">Paging queue type values</p><pre xml:space="preserve">
0 - VIDMM_PAGING_QUEUE_TYPE_UMD
1 - VIDMM_PAGING_QUEUE_TYPE_Default
2 - VIDMM_PAGING_QUEUE_TYPE_Evict
3 - VIDMM_PAGING_QUEUE_TYPE_Reclaim
   </pre><p class="p">Packet type values</p><pre xml:space="preserve">
0 - DXGKETW_RENDER_COMMAND_BUFFER
1 - DXGKETW_DEFERRED_COMMAND_BUFFER
2 - DXGKETW_SYSTEM_COMMAND_BUFFER
3 - DXGKETW_MMIOFLIP_COMMAND_BUFFER
4 - DXGKETW_WAIT_COMMAND_BUFFER
5 - DXGKETW_SIGNAL_COMMAND_BUFFER
6 - DXGKETW_DEVICE_COMMAND_BUFFER
7 - DXGKETW_SOFTWARE_COMMAND_BUFFER
   </pre><p class="p">Engine type values</p><pre xml:space="preserve">
0 - DXGK_ENGINE_TYPE_OTHER
1 - DXGK_ENGINE_TYPE_3D
2 - DXGK_ENGINE_TYPE_VIDEO_DECODE
3 - DXGK_ENGINE_TYPE_VIDEO_ENCODE
4 - DXGK_ENGINE_TYPE_VIDEO_PROCESSING
5 - DXGK_ENGINE_TYPE_SCENE_ASSEMBLY
6 - DXGK_ENGINE_TYPE_COPY
7 - DXGK_ENGINE_TYPE_OVERLAY
8 - DXGK_ENGINE_TYPE_CRYPTO
   </pre><p class="p">DMA interrupt type values</p><pre xml:space="preserve">
1 = DXGK_INTERRUPT_DMA_COMPLETED
2 = DXGK_INTERRUPT_DMA_PREEMPTED
4 = DXGK_INTERRUPT_DMA_FAULTED
9 = DXGK_INTERRUPT_DMA_PAGE_FAULTED
   </pre><p class="p">Queue type values</p><pre xml:space="preserve">
0 = Queue_Packet
1 = Dma_Packet
2 = Paging_Queue_Packet
   </pre><p class="p"><strong class="ph b">Driver Events</strong></p>
                        <p class="p">Load balance event type values</p><pre xml:space="preserve">
1 - LoadBalanceEvent_GPU
8 - LoadBalanceEvent_CPU
21 - LoadBalanceMasterEvent_GPU
22 - LoadBalanceMasterEvent_CPU
   </pre><p class="p"><strong class="ph b">OpenMP Events</strong></p>
                        <p class="p">OpenMP event class values</p><pre xml:space="preserve">
78 - TRACE_PROCESS_EVENT_OPENMP
79 - TRACE_PROCESS_EVENT_OPENMP_START
80 - TRACE_PROCESS_EVENT_OPENMP_FINISH
   </pre><p class="p">OpenMP event kind values</p><pre xml:space="preserve">
15 - OPENMP_EVENT_KIND_TASK_CREATE
16 - OPENMP_EVENT_KIND_TASK_SCHEDULE
17 - OPENMP_EVENT_KIND_CANCEL
20 - OPENMP_EVENT_KIND_MUTEX_RELEASED
21 - OPENMP_EVENT_KIND_LOCK_INIT
22 - OPENMP_EVENT_KIND_LOCK_DESTROY
25 - OPENMP_EVENT_KIND_DISPATCH
26 - OPENMP_EVENT_KIND_FLUSH
27 - OPENMP_EVENT_KIND_THREAD
28 - OPENMP_EVENT_KIND_PARALLEL
29 - OPENMP_EVENT_KIND_SYNC_REGION_WAIT
30 - OPENMP_EVENT_KIND_SYNC_REGION
31 - OPENMP_EVENT_KIND_TASK
32 - OPENMP_EVENT_KIND_MASTER
33 - OPENMP_EVENT_KIND_REDUCTION
34 - OPENMP_EVENT_KIND_MUTEX_WAIT
35 - OPENMP_EVENT_KIND_CRITICAL_SECTION
36 - OPENMP_EVENT_KIND_WORKSHARE
   </pre><p class="p">OpenMP thread type values</p><pre xml:space="preserve">
1 - OpenMP Initial Thread
2 - OpenMP Worker Thread
3 - OpenMP Internal Thread
4 - Unknown
   </pre><p class="p">OpenMP sync region kind values</p><pre xml:space="preserve">
1 - Barrier
2 - Implicit barrier
3 - Explicit barrier
4 - Implementation-dependent barrier
5 - Taskwait
6 - Taskgroup
    </pre><p class="p">OpenMP task kind values</p><pre xml:space="preserve">
1 - Initial task
2 - Implicit task
3 - Explicit task
    </pre><p class="p">OpenMP prior task status values</p><pre xml:space="preserve">
1 - Task completed
2 - Task yielded to another task
3 - Task was cancelled
7 - Task was switched out for other reasons
    </pre><p class="p">OpenMP mutex kind values</p><pre xml:space="preserve">
1 - Waiting for lock
2 - Testing lock
3 - Waiting for nested lock
4 - Tesing nested lock
5 - Waitng for entering critical section region
6 - Waiting for entering atomic region
7 - Waiting for entering ordered region
    </pre><p class="p">OpenMP critical section kind values</p><pre xml:space="preserve">
5 - Critical section region
6 - Atomic region
7 - Ordered region
    </pre><p class="p">OpenMP workshare kind values</p><pre xml:space="preserve">
1 - Loop region
2 - Sections region
3 - Single region (executor)
4 - Single region (waiting)
5 - Workshare region
6 - Distrubute region
7 - Taskloop region
    </pre><p class="p">OpenMP dispatch kind values</p><pre xml:space="preserve">
1 - Iteration
2 - Section
    </pre></div>
                  </div>
                  <div class="topic nested1" id="export-sqlite-examples"><a name="export-sqlite-examples" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#export-sqlite-examples" name="export-sqlite-examples" shape="rect">3.3.&nbsp;Common SQLite Examples</a></h3>
                     <div class="body">
                        <p class="p"><strong class="ph b">Common Helper Commands</strong></p>
                        <p class="p">When utilizing sqlite3 command line tool, it’s helpful to have data
                           printed as named columns, this can be done with:
                        </p><pre xml:space="preserve">
.mode column
.headers on
    </pre><p class="p">Default column width is determined by the data in the first row of
                           results. If this doesn’t work out well, you can specify widths manually.
                        </p><pre xml:space="preserve">.width 10 20 50</pre><p class="p"><strong class="ph b">Obtaining Sample Report</strong></p>
                        <p class="p">CLI interface of <span class="ph">Nsight Systems</span> was used to profile radixSortThrust CUDA
                           sample, then the resulting .nsys-rep file was exported using the <span class="ph">nsys</span> export.
                           
                        </p><pre xml:space="preserve">
<span class="ph">nsys</span> profile --trace=cuda,osrt radixSortThrust
<span class="ph">nsys</span> export --type sqlite report1.nsys-rep
    </pre><p class="p"><strong class="ph b">Serialized Process and Thread Identifiers</strong></p>
                        <p class="p"><span class="ph">Nsight Systems</span> stores identifiers where events originated in
                           serialized form. For events that have globalTid or globalPid fields exported,
                           use the following code to extract numeric TID and PID.
                        </p><pre xml:space="preserve">
SELECT globalTid / 0x1000000 % 0x1000000 AS PID, globalTid % 0x1000000 AS TID FROM TABLE_NAME;
    </pre><p class="p">Note: globalTid field includes both TID and PID values, while globalPid
                           only containes the PID value.
                        </p>
                        <p class="p"><strong class="ph b">Correlate CUDA Kernel Launches With CUDA API Kernel Launches</strong></p><pre xml:space="preserve">
ALTER TABLE CUPTI_ACTIVITY_KIND_RUNTIME ADD COLUMN name TEXT;
ALTER TABLE CUPTI_ACTIVITY_KIND_RUNTIME ADD COLUMN kernelName TEXT;

UPDATE CUPTI_ACTIVITY_KIND_RUNTIME SET kernelName =
    (SELECT value FROM StringIds
    JOIN CUPTI_ACTIVITY_KIND_KERNEL AS cuda_gpu
        ON cuda_gpu.shortName = StringIds.id
        AND CUPTI_ACTIVITY_KIND_RUNTIME.correlationId = cuda_gpu.correlationId);

UPDATE CUPTI_ACTIVITY_KIND_RUNTIME SET name =
    (SELECT value FROM StringIds WHERE nameId = StringIds.id);
    </pre><p class="p">Select 10 longest CUDA API ranges that resulted in kernel execution.</p><pre xml:space="preserve">    
SELECT name, kernelName, start, end FROM CUPTI_ACTIVITY_KIND_RUNTIME
    WHERE kernelName IS NOT NULL ORDER BY end - start LIMIT 10;
    </pre><p class="p">Results:</p><pre xml:space="preserve">
name                    kernelName               start       end       
----------------------  -----------------------  ----------  ----------
cudaLaunchKernel_v7000  RadixSortScanBinsKernel  658863435   658868490 
cudaLaunchKernel_v7000  RadixSortScanBinsKernel  609755015   609760075 
cudaLaunchKernel_v7000  RadixSortScanBinsKernel  632683286   632688349 
cudaLaunchKernel_v7000  RadixSortScanBinsKernel  606495356   606500439 
cudaLaunchKernel_v7000  RadixSortScanBinsKernel  603114486   603119586 
cudaLaunchKernel_v7000  RadixSortScanBinsKernel  802729785   802734906 
cudaLaunchKernel_v7000  RadixSortScanBinsKernel  593381170   593386294 
cudaLaunchKernel_v7000  RadixSortScanBinsKernel  658759955   658765090 
cudaLaunchKernel_v7000  RadixSortScanBinsKernel  681549917   681555059 
cudaLaunchKernel_v7000  RadixSortScanBinsKernel  717812527   717817671
    </pre><p class="p"><strong class="ph b">Remove Ranges Overlapping With Overhead</strong></p>
                        <p class="p">Use the this query to count CUDA API ranges overlapping with the overhead
                           ones.
                        </p>
                        <p class="p">Replace "SELECT COUNT(*)" with "DELETE" to remove such ranges.</p><pre xml:space="preserve">
SELECT COUNT(*) FROM CUPTI_ACTIVITY_KIND_RUNTIME WHERE rowid IN
(
    SELECT cuda.rowid
    FROM PROFILER_OVERHEAD as overhead
    INNER JOIN CUPTI_ACTIVITY_KIND_RUNTIME as cuda ON
    (cuda.start BETWEEN overhead.start and overhead.end)
    OR (cuda.end BETWEEN overhead.start and overhead.end)
    OR (cuda.start &lt; overhead.start AND cuda.end &gt; overhead.end)
);
    </pre><p class="p">Results:</p><pre xml:space="preserve">
COUNT(*)  
----------
1095      
    </pre><p class="p"><strong class="ph b">Find CUDA API Calls That Resulted in Original Graph Node Creation.</strong></p><pre xml:space="preserve">
SELECT graph.graphNodeId, api.start, graph.start as graphStart, api.end,
    api.globalTid, api.correlationId, api.globalTid,
    (SELECT value FROM StringIds where api.nameId == id) as name
FROM CUPTI_ACTIVITY_KIND_RUNTIME as api
JOIN
    (
        SELECT start, graphNodeId, globalTid from CUDA_GRAPH_EVENTS
        GROUP BY graphNodeId
        HAVING COUNT(originalGraphNodeId) = 0
    ) as graph
ON api.globalTid == graph.globalTid AND api.start &lt; graph.start AND api.end &gt; graph.start
ORDER BY graphNodeId;
    </pre><p class="p">Results:</p><pre xml:space="preserve">
graphNodeId  start       graphStart  end         globalTid        correlationId  globalTid        name                         
-----------  ----------  ----------  ----------  ---------------  -------------  ---------------  -----------------------------
1            584366518   584378040   584379102   281560221750233  109            281560221750233  cudaGraphAddMemcpyNode_v10000
2            584379402   584382428   584383139   281560221750233  110            281560221750233  cudaGraphAddMemsetNode_v10000
3            584390663   584395352   584396053   281560221750233  111            281560221750233  cudaGraphAddKernelNode_v10000
4            584396314   584397857   584398438   281560221750233  112            281560221750233  cudaGraphAddMemsetNode_v10000
5            584398759   584400311   584400812   281560221750233  113            281560221750233  cudaGraphAddKernelNode_v10000
6            584401083   584403047   584403527   281560221750233  114            281560221750233  cudaGraphAddMemcpyNode_v10000
7            584403928   584404920   584405491   281560221750233  115            281560221750233  cudaGraphAddHostNode_v10000  
29           632107852   632117921   632121407   281560221750233  144            281560221750233  cudaMemcpyAsync_v3020        
30           632122168   632125545   632127989   281560221750233  145            281560221750233  cudaMemsetAsync_v3020        
31           632131546   632133339   632135584   281560221750233  147            281560221750233  cudaMemsetAsync_v3020        
34           632162514   632167393   632169297   281560221750233  151            281560221750233  cudaMemcpyAsync_v3020        
35           632170068   632173334   632175388   281560221750233  152            281560221750233  cudaLaunchHostFunc_v10000    
   </pre><p class="p"><strong class="ph b">Backtraces for OSRT Ranges</strong></p>
                        <p class="p">Adding text columns makes results of the query below more human-readable.</p><pre xml:space="preserve">
ALTER TABLE OSRT_API ADD COLUMN name TEXT;
UPDATE OSRT_API SET name = (SELECT value FROM StringIds WHERE OSRT_API.nameId = StringIds.id);

ALTER TABLE OSRT_CALLCHAINS ADD COLUMN symbolName TEXT;
UPDATE OSRT_CALLCHAINS SET symbolName = (SELECT value FROM StringIds WHERE symbol = StringIds.id);

ALTER TABLE OSRT_CALLCHAINS ADD COLUMN moduleName TEXT;
UPDATE OSRT_CALLCHAINS SET moduleName = (SELECT value FROM StringIds WHERE module = StringIds.id);
   </pre><p class="p">Print backtrace of the longest OSRT range</p><pre xml:space="preserve">
SELECT globalTid / 0x1000000 % 0x1000000 AS PID, globalTid % 0x1000000 AS TID,
    start, end, name, callchainId, stackDepth, symbolName, moduleName
FROM OSRT_API LEFT JOIN OSRT_CALLCHAINS ON callchainId == OSRT_CALLCHAINS.id
WHERE OSRT_API.rowid IN (SELECT rowid FROM OSRT_API ORDER BY end - start DESC LIMIT 1)
ORDER BY stackDepth LIMIT 10;
   </pre><p class="p">Results:</p><pre xml:space="preserve">
PID         TID         start       end         name                    callchainId  stackDepth  symbolName                      moduleName                              
----------  ----------  ----------  ----------  ----------------------  -----------  ----------  ------------------------------  ----------------------------------------
19163       19176       360897690   860966851   pthread_cond_timedwait  88           0           pthread_cond_timedwait@GLIBC_2  /lib/x86_64-linux-gnu/libpthread-2.27.so
19163       19176       360897690   860966851   pthread_cond_timedwait  88           1           0x7fbc983b7227                  /usr/lib/x86_64-linux-gnu/libcuda.so.418
19163       19176       360897690   860966851   pthread_cond_timedwait  88           2           0x7fbc9835d5c7                  /usr/lib/x86_64-linux-gnu/libcuda.so.418
19163       19176       360897690   860966851   pthread_cond_timedwait  88           3           0x7fbc983b64a8                  /usr/lib/x86_64-linux-gnu/libcuda.so.418
19163       19176       360897690   860966851   pthread_cond_timedwait  88           4           start_thread                    /lib/x86_64-linux-gnu/libpthread-2.27.so
19163       19176       360897690   860966851   pthread_cond_timedwait  88           5           __clone                         /lib/x86_64-linux-gnu/libc-2.27.so      
   </pre><p class="p">Profiled processes output streams</p><pre xml:space="preserve">
ALTER TABLE ProcessStreams ADD COLUMN filename TEXT;
UPDATE ProcessStreams SET filename = (SELECT value FROM StringIds WHERE ProcessStreams.filenameId = StringIds.id);

ALTER TABLE ProcessStreams ADD COLUMN content TEXT;
UPDATE ProcessStreams SET content = (SELECT value FROM StringIds WHERE ProcessStreams.contentId = StringIds.id);
   </pre><p class="p">Select all collected stdout and stderr streams.</p><pre xml:space="preserve">
select globalPid / 0x1000000 % 0x1000000 AS PID, filename, content from ProcessStreams;
   </pre><p class="p">Results:</p><pre xml:space="preserve">
PID         filename                                                 content                                                                                                                                                                                                                                                                                                                    
----------  -------------------------------------------------------  --------------------------------------------------------------------------------------------------------------------                                                                                                                                                                                                       
19163       /tmp/nvidia/nsight_systems/streams/pid_19163_stdout.log  /home/user_name/NVIDIA_CUDA-10.1_Samples/6_Advanced/radixSortThrust/radixSortThrust Starting...

GPU Device 0: "Quadro P2000" with compute capability 6.1


Sorting 1048576 32-bit unsigned int keys and values

radixSortThrust, Throughput = 401.0872 MElements/s, Time = 0.00261 s, Size = 1048576 elements
Test passed

19163       /tmp/nvidia/nsight_systems/streams/pid_19163_stderr.log
   </pre><p class="p"><strong class="ph b">Thread Summary</strong></p>
                        <p class="p">Please note, that <span class="ph">Nsight Systems</span> applies additional logic
                           during sampling events processing to work around lost events. This means that
                           the results of the below query might differ slightly from the ones shown in
                           “Analysis summary” tab.
                        </p>
                        <p class="p">Thread summary calculated using CPU cycles (when available).</p><pre xml:space="preserve">
SELECT
    globalTid / 0x1000000 % 0x1000000 AS PID,
    globalTid % 0x1000000 AS TID,
    ROUND(100.0 * SUM(cpuCycles) / 
        (
            SELECT SUM(cpuCycles) FROM COMPOSITE_EVENTS 
            GROUP BY globalTid / 0x1000000000000 % 0x100
        ),
        2
    ) as CPU_utilization,
    (SELECT value FROM StringIds WHERE id = 
        (
            SELECT nameId FROM ThreadNames 
            WHERE ThreadNames.globalTid = COMPOSITE_EVENTS.globalTid
        )
    ) as thread_name
FROM COMPOSITE_EVENTS
GROUP BY globalTid
ORDER BY CPU_utilization DESC
LIMIT 10;
   </pre><p class="p">Results:</p><pre xml:space="preserve">
PID         TID         CPU_utilization  thread_name    
----------  ----------  ---------------  ---------------
19163       19163       98.4             radixSortThrust
19163       19168       1.35             CUPTI worker th
19163       19166       0.25             [NS]    
   </pre><p class="p">Thread running time may be calculated using scheduling data, when PMU
                           counter data was not collected.
                        </p><pre xml:space="preserve">
CREATE INDEX sched_start ON SCHED_EVENTS (start);

CREATE TABLE CPU_USAGE AS
SELECT
    first.globalTid as globalTid,
    (SELECT nameId FROM ThreadNames WHERE ThreadNames.globalTid = first.globalTid) as nameId,
    sum(second.start - first.start) as total_duration,
    count() as ranges_count
FROM SCHED_EVENTS as first
LEFT JOIN SCHED_EVENTS as second
ON second.rowid =
    (
        SELECT rowid
        FROM SCHED_EVENTS
        WHERE start &gt; first.start AND globalTid = first.globalTid
        ORDER BY start ASC
        LIMIT 1
    )
WHERE first.isSchedIn != 0
GROUP BY first.globalTid
ORDER BY total_duration DESC;

SELECT
    globalTid / 0x1000000 % 0x1000000 AS PID,
    globalTid % 0x1000000 AS TID,
    (SELECT value FROM StringIds where nameId == id) as thread_name,
    ROUND(100.0 * total_duration / (SELECT SUM(total_duration) FROM CPU_USAGE), 2) as CPU_utilization
FROM CPU_USAGE
ORDER BY CPU_utilization DESC;
   </pre><p class="p">Results:</p><pre xml:space="preserve">
PID         TID         thread_name      CPU_utilization
----------  ----------  ---------------  ---------------
19163       19163       radixSortThrust  93.74          
19163       19169       radixSortThrust  3.22           
19163       19168       CUPTI worker th  2.46           
19163       19166       [NS]             0.44           
19163       19172       radixSortThrust  0.07           
19163       19167       [NS Comms]       0.05           
19163       19176       radixSortThrust  0.02           
19163       19170       radixSortThrust  0.0
   </pre><p class="p"><strong class="ph b">Function Table</strong></p>
                        <p class="p">These examples demonstrate how to calculate Flat and BottomUp (for top
                           level only) views statistics.
                        </p>
                        <p class="p">To set up:</p><pre xml:space="preserve">
ALTER TABLE SAMPLING_CALLCHAINS ADD COLUMN symbolName TEXT;
UPDATE SAMPLING_CALLCHAINS SET symbolName = (SELECT value FROM StringIds WHERE symbol = StringIds.id);

ALTER TABLE SAMPLING_CALLCHAINS ADD COLUMN moduleName TEXT;
UPDATE SAMPLING_CALLCHAINS SET moduleName = (SELECT value FROM StringIds WHERE module = StringIds.id);
   </pre><p class="p">To get flat view:</p><pre xml:space="preserve">
SELECT symbolName, moduleName, ROUND(100.0 * sum(cpuCycles) /
    (SELECT SUM(cpuCycles) FROM COMPOSITE_EVENTS), 2) AS flatTimePercentage
FROM SAMPLING_CALLCHAINS
LEFT JOIN COMPOSITE_EVENTS ON SAMPLING_CALLCHAINS.id == COMPOSITE_EVENTS.id
GROUP BY symbol, module
ORDER BY flatTimePercentage DESC
LIMIT 5;
   </pre><p class="p">To get BottomUp view (top level only):</p><pre xml:space="preserve">
SELECT symbolName, moduleName, ROUND(100.0 * sum(cpuCycles) /
    (SELECT SUM(cpuCycles) FROM COMPOSITE_EVENTS), 2) AS selfTimePercentage
FROM SAMPLING_CALLCHAINS
LEFT JOIN COMPOSITE_EVENTS ON SAMPLING_CALLCHAINS.id == COMPOSITE_EVENTS.id
WHERE stackDepth == 0
GROUP BY symbol, module
ORDER BY selfTimePercentage DESC
LIMIT 5;
   </pre><p class="p">Results:</p><pre xml:space="preserve">
symbolName   moduleName   flatTimePercentage
-----------  -----------  ------------------
[Max depth]  [Max depth]  99.92             
thrust::zip  /home/user_  24.17             
thrust::zip  /home/user_  24.17             
thrust::det  /home/user_  24.17             
thrust::det  /home/user_  24.17             
symbolName      moduleName                                   selfTimePercentage
--------------  -------------------------------------------  ------------------
0x7fbc984982b6  /usr/lib/x86_64-linux-gnu/libcuda.so.418.39  5.29              
0x7fbc982d0010  /usr/lib/x86_64-linux-gnu/libcuda.so.418.39  2.81              
thrust::iterat  /home/user_name/NVIDIA_CUDA-10.1_Samples/6_  2.23              
thrust::iterat  /home/user_name/NVIDIA_CUDA-10.1_Samples/6_  1.55              
void thrust::i  /home/user_name/NVIDIA_CUDA-10.1_Samples/6_  1.55
   </pre><p class="p"><strong class="ph b">DX12 API Frame Duration Histogram</strong></p>
                        <p class="p">The example demonstrates how to calculate DX12 CPU frames durartion and
                           construct a histogram out of it.
                        </p><pre xml:space="preserve">
CREATE INDEX DX12_API_ENDTS ON DX12_API (end);

CREATE TEMP VIEW DX12_API_FPS AS SELECT end AS start,
    (SELECT end FROM DX12_API
        WHERE end &gt; outer.end AND nameId == (SELECT id FROM StringIds
            WHERE value == "IDXGISwapChain::Present")
        ORDER BY end ASC LIMIT 1) AS end
FROM DX12_API AS outer
    WHERE nameId == (SELECT id FROM StringIds WHERE value == "IDXGISwapChain::Present")
ORDER BY end;
   </pre><p class="p">Number of frames with a duration of [X, X + 1) milliseconds.</p><pre xml:space="preserve">
SELECT
    CAST((end - start) / 1000000.0 AS INT) AS duration_ms,
    count(*)
FROM DX12_API_FPS
WHERE end IS NOT NULL
GROUP BY duration_ms
ORDER BY duration_ms;
   </pre><p class="p">Results:</p><pre xml:space="preserve">
duration_ms  count(*)  
-----------  ----------
3            1         
4            2         
5            7         
6            153       
7            19        
8            116       
9            16        
10           8         
11           2         
12           2         
13           1         
14           4         
16           3         
17           2         
18           1
   </pre><p class="p"><strong class="ph b">GPU Context Switch Events Enumeration</strong></p>
                        <p class="p">GPU context duration is between first BEGIN and a matching END event.</p><pre xml:space="preserve">
SELECT (CASE tag WHEN 8 THEN "BEGIN" WHEN 7 THEN "END" END) AS tag,
    globalPid / 0x1000000 % 0x1000000 AS PID,
    vmId, seqNo, contextId, timestamp, gpuId FROM FECS_EVENTS
WHERE tag in (7, 8) ORDER BY seqNo LIMIT 10;
   </pre><p class="p">Results:</p><pre xml:space="preserve">
tag         PID         vmId        seqNo       contextId   timestamp   gpuId     
----------  ----------  ----------  ----------  ----------  ----------  ----------
BEGIN       23371       0           0           1048578     56759171    0         
BEGIN       23371       0           1           1048578     56927765    0         
BEGIN       23371       0           3           1048578     63799379    0         
END         23371       0           4           1048578     63918806    0         
BEGIN       19397       0           5           1048577     64014692    0         
BEGIN       19397       0           6           1048577     64250369    0         
BEGIN       19397       0           8           1048577     1918310004  0         
END         19397       0           9           1048577     1918521098  0         
BEGIN       19397       0           10          1048577     2024164744  0         
BEGIN       19397       0           11          1048577     2024358650  0   
   </pre><p class="p"><strong class="ph b">Resolve NVTX Category Name</strong></p>
                        <p class="p">The example demonstrates how to resolve NVTX category name for NVTX marks
                           and ranges.
                        </p><pre xml:space="preserve">
WITH
  event AS (
    SELECT *
    FROM NVTX_EVENTS
    WHERE eventType IN (34, 59, 60) -- mark, push/pop, start/end
  ),
  category AS (
    SELECT
      category,
      domainId,
      text AS categoryName
    FROM NVTX_EVENTS
    WHERE eventType == 33 --  new category
  )
SELECT
  start,
  end,
  globalTid,
  eventType,
  domainId,
  category,
  categoryName,
  text
FROM event JOIN category USING (category, domainId)
ORDER BY start;
   </pre><p class="p">Results:</p><pre xml:space="preserve">
start       end         globalTid        eventType   domainId    category    categoryName               text            
----------  ----------  ---------------  ----------  ----------  ----------  -------------------------  ----------------
18281150    18311960    281534938484214  59          0           1           FirstCategoryUnderDefault  Push Pop Range A
18288187    18306674    281534938484214  59          0           2           SecondCategoryUnderDefaul  Push Pop Range B
18294247                281534938484214  34          0           1           FirstCategoryUnderDefault  Mark A          
18300034                281534938484214  34          0           2           SecondCategoryUnderDefaul  Mark B          
18345546    18372595    281534938484214  60          1           1           FirstCategoryUnderMyDomai  Start End Range 
18352924    18378342    281534938484214  60          1           2           SecondCategoryUnderMyDoma  Start End Range 
18359634                281534938484214  34          1           1           FirstCategoryUnderMyDomai  Mark A          
18365448                281534938484214  34          1           2           SecondCategoryUnderMyDoma  Mark B 
   </pre><p class="p"><strong class="ph b">Rename CUDA Kernels with NVTX</strong></p>
                        <p class="p">The example demonstrates how to map innermost NVTX push-pop range to a
                           matching CUDA kernel run.
                        </p><pre xml:space="preserve">
ALTER TABLE CUPTI_ACTIVITY_KIND_KERNEL ADD COLUMN nvtxRange TEXT;
CREATE INDEX nvtx_start ON NVTX_EVENTS (start);


UPDATE CUPTI_ACTIVITY_KIND_KERNEL SET nvtxRange = (
    SELECT NVTX_EVENTS.text
    FROM NVTX_EVENTS JOIN CUPTI_ACTIVITY_KIND_RUNTIME ON
        NVTX_EVENTS.eventType == 59 AND
        NVTX_EVENTS.globalTid == CUPTI_ACTIVITY_KIND_RUNTIME.globalTid AND
        NVTX_EVENTS.start &lt;= CUPTI_ACTIVITY_KIND_RUNTIME.start AND
        NVTX_EVENTS.end &gt;= CUPTI_ACTIVITY_KIND_RUNTIME.end
    WHERE
        CUPTI_ACTIVITY_KIND_KERNEL.correlationId == CUPTI_ACTIVITY_KIND_RUNTIME.correlationId
    ORDER BY NVTX_EVENTS.start DESC LIMIT 1
);

SELECT start, end, globalPid, StringIds.value as shortName, nvtxRange
FROM CUPTI_ACTIVITY_KIND_KERNEL JOIN StringIds ON shortName == id
ORDER BY start LIMIT 6;
   </pre><p class="p">Results:</p><pre xml:space="preserve">
start       end         globalPid          shortName      nvtxRange 
----------  ----------  -----------------  -------------  ----------
526545376   526676256   72057700439031808  MatrixMulCUDA            
526899648   527030368   72057700439031808  MatrixMulCUDA  Add       
527031648   527162272   72057700439031808  MatrixMulCUDA  Add       
527163584   527294176   72057700439031808  MatrixMulCUDA  My Kernel 
527296160   527426592   72057700439031808  MatrixMulCUDA  My Range  
527428096   527558656   72057700439031808  MatrixMulCUDA   
   </pre><p class="p"><strong class="ph b">Select CUDA Calls With Backtraces</strong></p><pre xml:space="preserve">
ALTER TABLE CUPTI_ACTIVITY_KIND_RUNTIME ADD COLUMN name TEXT;
UPDATE CUPTI_ACTIVITY_KIND_RUNTIME SET name = (SELECT value FROM StringIds WHERE CUPTI_ACTIVITY_KIND_RUNTIME.nameId = StringIds.id);

ALTER TABLE CUDA_CALLCHAINS ADD COLUMN symbolName TEXT;
UPDATE CUDA_CALLCHAINS SET symbolName = (SELECT value FROM StringIds WHERE symbol = StringIds.id);

SELECT globalTid % 0x1000000 AS TID,
    start, end, name, callchainId, stackDepth, symbolName
FROM CUDA_CALLCHAINS JOIN CUPTI_ACTIVITY_KIND_RUNTIME ON callchainId == CUDA_CALLCHAINS.id
ORDER BY callchainId, stackDepth LIMIT 11;
   </pre><p class="p">Results:</p><pre xml:space="preserve">
TID         start       end         name           callchainId  stackDepth  symbolName    
----------  ----------  ----------  -------------  -----------  ----------  --------------
11928       168976467   169077826   cuMemAlloc_v2  1            0           0x7f13c44f02ab
11928       168976467   169077826   cuMemAlloc_v2  1            1           0x7f13c44f0b8f
11928       168976467   169077826   cuMemAlloc_v2  1            2           0x7f13c44f3719
11928       168976467   169077826   cuMemAlloc_v2  1            3           cuMemAlloc_v2 
11928       168976467   169077826   cuMemAlloc_v2  1            4           cudart::driver
11928       168976467   169077826   cuMemAlloc_v2  1            5           cudart::cudaAp
11928       168976467   169077826   cuMemAlloc_v2  1            6           cudaMalloc    
11928       168976467   169077826   cuMemAlloc_v2  1            7           cudaError cuda
11928       168976467   169077826   cuMemAlloc_v2  1            8           main          
11928       168976467   169077826   cuMemAlloc_v2  1            9           __libc_start_m
11928       168976467   169077826   cuMemAlloc_v2  1            10          _start  
   </pre><p class="p"><strong class="ph b">SLI Peer-to-Peer Query</strong></p>
                        <p class="p">The example demonstrates how to query SLI Peer-to-Peer events with resource
                           size greater than value and within a time range sorted by resource size
                           descending.
                        </p><pre xml:space="preserve">
SELECT *
FROM SLI_P2P
WHERE resourceSize &lt; 98304 AND start &gt; 1568063100 AND end &lt; 1579468901
ORDER BY resourceSize DESC;
   </pre><p class="p">Results:</p><pre xml:space="preserve">
start       end         eventClass  globalTid          gpu         frameId     transferSkipped  srcGpu      dstGpu      numSubResources  resourceSize  subResourceIdx  smplWidth   smplHeight  smplDepth   bytesPerElement  dxgiFormat  logSurfaceNames  transferInfo  isEarlyPushManagedByNvApi  useAsyncP2pForResolve  transferFuncName  regimeName  debugName   bindType  
----------  ----------  ----------  -----------------  ----------  ----------  ---------------  ----------  ----------  ---------------  ------------  --------------  ----------  ----------  ----------  ---------------  ----------  ---------------  ------------  -------------------------  ---------------------  ----------------  ----------  ----------  ----------
1570351100  1570351101  62          72057698056667136  0           771         0                256         512         1                1048576       0               256         256         1           16               2                            3             0                          0                                                                          
1570379300  1570379301  62          72057698056667136  0           771         0                256         512         1                1048576       0               64          64          64          4                31                           3             0                          0                                                                          
1572316400  1572316401  62          72057698056667136  0           773         0                256         512         1                1048576       0               256         256         1           16               2                            3             0                          0                                                                          
1572345400  1572345401  62          72057698056667136  0           773         0                256         512         1                1048576       0               64          64          64          4                31                           3             0                          0                                                                          
1574734300  1574734301  62          72057698056667136  0           775         0                256         512         1                1048576       0               256         256         1           16               2                            3             0                          0                                                                          
1574767200  1574767201  62          72057698056667136  0           775         0                256         512         1                1048576       0               64          64          64          4                31                           3             0                          0                                                                          
   </pre><p class="p"><strong class="ph b">Generic Events</strong></p>
                        <p class="p">Syscall usage histogram by PID:</p><pre xml:space="preserve">
SELECT json_extract(data, '$.common_pid') AS PID, count(*) AS total
FROM GENERIC_EVENTS WHERE PID IS NOT NULL AND typeId = (
  SELECT typeId FROM GENERIC_EVENT_TYPES
  WHERE json_extract(data, '$.Name') = "raw_syscalls:sys_enter")
GROUP BY PID
ORDER BY total DESC
LIMIT 10;
   </pre><p class="p">Results:</p><pre xml:space="preserve">
PID         total     
----------  ----------
5551        32811     
9680        3988      
4328        1477      
9564        1246      
4376        1204      
4377        1167      
4357        656       
4355        655       
4356        640       
4354        633
   </pre><p class="p"><strong class="ph b">Fetching Generic Events in JSON Format</strong></p>
                        <p class="p">Text and JSON export modes don’t include generic events. Use the below
                           queries (without LIMIT clause) to extract JSON lines representation of generic
                           events, types and sources.
                        </p><pre xml:space="preserve">
SELECT json_insert('{}',
    '$.sourceId', sourceId,
    '$.data', json(data)
)
FROM GENERIC_EVENT_SOURCES LIMIT 2;

SELECT json_insert('{}',
    '$.typeId', typeId,
    '$.sourceId', sourceId,
    '$.data', json(data)
)
FROM GENERIC_EVENT_TYPES LIMIT 2;

SELECT json_insert('{}',
    '$.rawTimestamp', rawTimestamp,
    '$.timestamp', timestamp,
    '$.typeId', typeId,
    '$.data', json(data)
)
FROM GENERIC_EVENTS LIMIT 2;
   </pre><p class="p">Results:</p><pre xml:space="preserve">
json_insert('{}',
    '$.sourceId', sourceId,
    '$.data', json(data)
)                                       
---------------------------------------------------------------------------------------------------------------
{"sourceId":72057602627862528,"data":{"Name":"FTrace","TimeSource":"ClockMonotonicRaw","SourceGroup":"FTrace"}}
json_insert('{}',
    '$.typeId', typeId,
    '$.sourceId', sourceId,
    '$.data', json(data)
)                                                                                                                                                                                                                                             
--------------------------------------------------------------------------------------------------------------------                                                                                                                                                                                                                                                                                                                                                                                                                                                                                             
{"typeId":72057602627862547,"sourceId":72057602627862528,"data":{"Name":"raw_syscalls:sys_enter","Format":"\"NR %ld (%lx, %lx, %lx, %lx, %lx, %lx)\", REC-&gt;id, REC-&gt;args[0], REC-&gt;args[1], REC-&gt;args[2], REC-&gt;args[3], REC-&gt;args[4], REC-&gt;args[5]","Fields":[{"Name":"common_pid","Prefix":"int","Suffix":""},{"Name":"id","Prefix":"long","S
{"typeId":72057602627862670,"sourceId":72057602627862528,"data":{"Name":"irq:irq_handler_entry","Format":"\"irq=%d name=%s\", REC-&gt;irq, __get_str(name)","Fields":[{"Name":"common_pid","Prefix":"int","Suffix":""},{"Name":"irq","Prefix":"int","Suffix":""},{"Name":"name","Prefix":"__data_loc char[]","Suffix":""},{"Name":"common_type",
json_insert('{}',
    '$.rawTimestamp', rawTimestamp,
    '$.timestamp', timestamp,
    '$.typeId', typeId,
    '$.data', json(data)
)                                                                 
--------------------------------------------------------------------------------------------------------------------                                                                                   
{"rawTimestamp":1183694330725221,"timestamp":6236683,"typeId":72057602627862670,"data":{"common_pid":"0","irq":"66","name":"327696","common_type":"142","common_flags":"9","common_preempt_count":"0"}}
{"rawTimestamp":1183694333695687,"timestamp":9207149,"typeId":72057602627862670,"data":{"common_pid":"0","irq":"66","name":"327696","common_type":"142","common_flags":"9","common_preempt_count":"0"}}


</pre></div>
                  </div>
                  <div class="topic nested1" id="exporter-arrow"><a name="exporter-arrow" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#exporter-arrow" name="exporter-arrow" shape="rect">Arrow Format Description</a></h3>
                     <div class="body">
                        <p class="p">
                           The Arrow type exported file uses the IPC stream format to store
                           the data in a file. The tables can be read by opening the file as
                           an arrow stream. For example one can use the
                           <samp class="ph codeph">open_stream</samp> function from the arrow python
                           package. For more information on the interfaces that can be used to
                           read an IPC stream file, please refer to the Apache Arrow
                           documentation
                           [<a class="xref" href="https://arrow.apache.org/docs/python/api/ipc.html" target="_blank" shape="rect">1</a>,
                           <a class="xref" href="https://arrow.apache.org/docs/python/ipc.html" target="_blank" shape="rect">2</a>].
                           
                        </p>
                        <p class="p">
                           The name of each table is included in the schema metadata. Thus,
                           while reading each table, the user can extract the table title from
                           the metadata. The table name metadata field has the key
                           <samp class="ph codeph">table_name</samp>. The titles of all the available tables
                           can be found in section
                           <a class="xref" href="index.html#exporter-sqlite-schema" shape="rect">SQLite Schema Reference</a>.
                           
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="exporter-json-txt"><a name="exporter-json-txt" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#exporter-json-txt" name="exporter-json-txt" shape="rect">JSON and Text Format Description</a></h3>
                     <div class="body">
                        <p class="p">JSON and TXT export formats are generated by serializing buffered messages,
                           	each on a new line. First, all collected events are processed. Then strings
                           	are serialized, followed by stdout, stderr streams if any, followed by thread
                           	names.
                        </p>
                        <p class="p">Output layout:</p><pre xml:space="preserve">
{Event #1}
{Event #2}
...
{Event #N}
{Strings}
{Streams}
{Threads}
    </pre><p class="p">For easier grepping of JSON output, the <samp class="ph codeph">--separate-strings</samp>
                           	switch may be used to force manual splitting of strings, streams and thread
                           	names data.
                        </p>
                        <p class="p">Example line split: <samp class="ph codeph">nsys export --export-json --separate-strings
                              	sample.nsys-rep -- -</samp></p><pre xml:space="preserve">
{"type":"String","id":"3720","value":"Process 14944 was launched by the profiler"}
{"type":"String","id":"3721","value":"Profiling has started."}
{"type":"String","id":"3722","value":"Profiler attached to the process."}
{"type":"String","id":"3723","value":"Profiling has stopped."}
{"type":"ThreadName","globalTid":"72057844756653436","nameId":"14","priority":"10"}
{"type":"ThreadName","globalTid":"72057844756657940","nameId":"15","priority":"10"}
{"type":"ThreadName","globalTid":"72057844756654400","nameId":"24","priority":"10"}
    </pre><p class="p">Compare with: <samp class="ph codeph">nsys export --export-json sample.nsys-rep -- -</samp></p><pre xml:space="preserve">
{"data":["[Unknown]","[Unknown kernel module]","[Max depth]","[Broken backtraces]",
  "[Called from Java]","QnxKernelTrace","mm_","task_submit","class_id","syncpt_id",
  "syncpt_thresh","pid","tid","FTrace","[NSys]","[NSys Comms]", "..." ,"Process
  14944 was launched by the profiler","Profiling has started.","Profiler attached
  to the process.","Profiling has stopped."]}
{"data":[{"nameIdx":"14","priority":"10","globalTid":"72057844756653436"},
  {"nameIdx":"15","priority":"10","globalTid":"72057844756657940"},{"nameIdx":"24",
  "priority":"10","globalTid":"72057844756654400"}]}
    </pre><p class="p">Note, that only last few lines are shown here for clarity and that carriage
                           	returns and indents were added to avoid wrapping documentation.
                        </p>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="report-scripts"><a name="report-scripts" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#report-scripts" name="report-scripts" shape="rect">4.&nbsp;Report Scripts</a></h2>
                  <div class="body"></div>
                  <div class="topic nested1" id="existing-report-scripts"><a name="existing-report-scripts" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#existing-report-scripts" name="existing-report-scripts" shape="rect">Report Scripts Shipped With Nsight Systems</a></h3>
                     <div class="body">
                        <p class="p">
                           The <span class="ph">Nsight Systems</span> development team created and
                           maintains a set of report scripts for some of the commonly requested
                           reports. These scripts will be updated to adapt to any changes in SQLite
                           schema or internal data structures.
                        </p>
                        <p class="p">
                           These scripts are located in the <span class="ph">Nsight Systems</span> package in
                           the Target-&lt;architecture&gt;/reports directory. The following standard
                           reports are available:
                        </p>
                     </div>
                     <div class="topic nested2" id="apigpusum"><a name="apigpusum" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#apigpusum" name="apigpusum" shape="rect">apigpusum[:base] -- CUDA API &amp; GPU Summary (CUDA API + kernels
                              + memory ops)</a></h3>
                        <div class="body">
                           <p class="p">Arguments</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">base</strong> - Optional argument, if given, will cause summary to
                                 be over the base name of the kernel, rather than the templated name.
                              </li>
                           </ul>
                           <p class="p">Output: All time values given in nanoseconds</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">Time(%)</strong> : Percentage of <strong class="ph b">Total Time</strong></li>
                              <li class="li"><strong class="ph b">Total Time</strong> : The total time used by all executions of this
                                 kernel
                              </li>
                              <li class="li"><strong class="ph b">Instances</strong>: The number of executions of this object
                              </li>
                              <li class="li"><strong class="ph b">Average</strong> : The average execution time of this kernel
                              </li>
                              <li class="li"><strong class="ph b">Minimum</strong> : The smallest execution time of this kernel
                              </li>
                              <li class="li"><strong class="ph b">Maximum</strong> : The largest execution time of this kernel
                              </li>
                              <li class="li"><strong class="ph b">Category</strong> : The category of the operation
                              </li>
                              <li class="li"><strong class="ph b">Operation</strong> : The name of the kernel
                              </li>
                           </ul>
                           <p class="p">
                              This report provides a summary of CUDA API calls, kernels and memory
                              operations, and their execution times. Note that the <strong class="ph b">Time(%)</strong>
                              column is calculated using a summation of the <strong class="ph b">Total Time</strong> column,
                              and represents that API call's, kernel's, or memory operation's
                              percent of the execution time of the APIs, kernels and memory
                              operations listed, and not a percentage of the application wall or CPU
                              execution time.
                           </p>
                           <p class="p">This report combines data from the <samp class="ph codeph">cudaapisum</samp>,
                              <samp class="ph codeph">gpukernsum</samp>, and <samp class="ph codeph">gpumemsizesum</samp> reports.
                              It is very similar to profile section of
                              <samp class="ph codeph">nvprof --dependency-analysis</samp>.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="cudaapisum"><a name="cudaapisum" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cudaapisum" name="cudaapisum" shape="rect">cudaapisum -- CUDA API Summary</a></h3>
                        <div class="body">
                           <p class="p">Arguments - None</p>
                           <p class="p">Output: All time values given in nanoseconds</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">Time(%)</strong> : Percentage of <strong class="ph b">Total Time</strong></li>
                              <li class="li"><strong class="ph b">Total Time</strong> : The total time used by all executions of this
                                 function
                              </li>
                              <li class="li"><strong class="ph b">Num Calls</strong> : The number of calls to this function
                              </li>
                              <li class="li"><strong class="ph b">Average</strong> : The average execution time of this function
                              </li>
                              <li class="li"><strong class="ph b">Minimum</strong> : The smallest execution time of this function
                              </li>
                              <li class="li"><strong class="ph b">Maximum</strong> : The largest execution time of this function
                              </li>
                              <li class="li"><strong class="ph b">Name</strong> : The name of the function
                              </li>
                           </ul>
                           <p class="p">
                              This report provides a summary of CUDA API functions and their
                              execution times. Note that the <strong class="ph b">Time(%)</strong> column is calculated
                              using a summation of the <strong class="ph b">Total Time</strong> column, and represents that
                              function's percent of the execution time of the functions listed,
                              and not a percentage of the application wall or CPU execution time.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="cudaapitrace"><a name="cudaapitrace" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#cudaapitrace" name="cudaapitrace" shape="rect">cudaapitrace -- CUDA API Trace</a></h3>
                        <div class="body">
                           <p class="p">Arguments - None</p>
                           <p class="p">Output: All time values given in nanoseconds</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">Start</strong> : Timestamp when API call was made
                              </li>
                              <li class="li"><strong class="ph b">Duration</strong> : Length of API calls
                              </li>
                              <li class="li"><strong class="ph b">Name</strong> : API function name
                              </li>
                              <li class="li"><strong class="ph b">Result</strong> : return value of API call
                              </li>
                              <li class="li"><strong class="ph b">CorrID</strong> : Correlation used to map to other CUDA calls
                              </li>
                              <li class="li"><strong class="ph b">Pid</strong> : Process ID that made the call
                              </li>
                              <li class="li"><strong class="ph b">Tid</strong> : Thread ID that made the call
                              </li>
                              <li class="li"><strong class="ph b">T-Pri</strong> : Run priority of call thread
                              </li>
                              <li class="li"><strong class="ph b">Thread Name</strong> : Name of thread that called API function
                              </li>
                           </ul>
                           <p class="p">This report provides a trace record of CUDA API function calls and
                              their execution times.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="gpukernsum"><a name="gpukernsum" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#gpukernsum" name="gpukernsum" shape="rect">gpukernsum[:base] -- CUDA GPU Kernel Summary</a></h3>
                        <div class="body">
                           <p class="p">Arguments</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">base</strong> - Optional argument, if given, will cause summary to
                                 be over the base name of the kernel, rather than the templated name.
                              </li>
                           </ul>
                           <p class="p">Output: All time values given in nanoseconds</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">Time(%)</strong> : Percentage of <strong class="ph b">Total Time</strong></li>
                              <li class="li"><strong class="ph b">Total Time</strong> : The total time used by all executions of this
                                 kernel
                              </li>
                              <li class="li"><strong class="ph b">Instances</strong> : The number of calls to this kernel
                              </li>
                              <li class="li"><strong class="ph b">Average</strong> : The average execution time of this kernel
                              </li>
                              <li class="li"><strong class="ph b">Minimum</strong> : The smallest execution time of this kernel
                              </li>
                              <li class="li"><strong class="ph b">Maximum</strong> : The largest execution time of this kernel
                              </li>
                              <li class="li"><strong class="ph b">Name</strong> : The name of the kernel
                              </li>
                           </ul>
                           <p class="p">This report provides a summary of CUDA kernels and their
                              execution times. Note that the <strong class="ph b">Time(%)</strong> column is calculated
                              using a summation of the <strong class="ph b">Total Time</strong> column, and represents that
                              kernel's percent of the execution time of the kernels listed, and not
                              a percentage of the application wall or CPU execution time.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="gpumemsizesum"><a name="gpumemsizesum" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#gpumemsizesum" name="gpumemsizesum" shape="rect">gpumemsizesum -- GPU Memory Operations Summary (by Size)</a></h3>
                        <div class="body">
                           <p class="p">Arguments - None</p>
                           <p class="p">Output: All memory values given in KiB</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">Total</strong> : Total number of KiB utilized by this operation
                              </li>
                              <li class="li"><strong class="ph b">Operations</strong> : Number of executions of this operation
                              </li>
                              <li class="li"><strong class="ph b">Average</strong> : The average memory size of this operation
                              </li>
                              <li class="li"><strong class="ph b">Minimum</strong> : The smallest memory size of this operation
                              </li>
                              <li class="li"><strong class="ph b">Maximum</strong> : The largest memory size of this operation
                              </li>
                              <li class="li"><strong class="ph b">Name</strong> : The name of the operation
                              </li>
                           </ul>
                           <p class="p">
                              This report provides a summary of GPU memory operations and the amount
                              of memory they utilize.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="gpumemtimesum"><a name="gpumemtimesum" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#gpumemtimesum" name="gpumemtimesum" shape="rect">gpumemtimesum -- GPU Memory Operations Summary (by Time)</a></h3>
                        <div class="body">
                           <p class="p">Arguments - None</p>
                           <p class="p">Output: All memory values given in KiB</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">Time(%)</strong> : Percentage of <strong class="ph b">Total Time</strong></li>
                              <li class="li"><strong class="ph b">Total Time</strong> : The total time used by all executions of this
                                 operation
                              </li>
                              <li class="li"><strong class="ph b">Operations</strong>: The number of operations of this type
                              </li>
                              <li class="li"><strong class="ph b">Average</strong> : The average execution time of this operation
                              </li>
                              <li class="li"><strong class="ph b">Minimum</strong> : The smallest execution time of this operation
                              </li>
                              <li class="li"><strong class="ph b">Maximum</strong> : The largest execution time of this operation
                              </li>
                              <li class="li"><strong class="ph b">Operation</strong> : The name of the memory operation
                              </li>
                           </ul>
                           <p class="p">
                              This report provides a summary of GPU memory operations and
                              their execution times. Note that the <strong class="ph b">Time(%)</strong> column is
                              calculated using a summation of the <strong class="ph b">Total Time</strong> column, and
                              represents that operation's percent of the execution time of the
                              operations listed, and not a percentage of the application wall or CPU
                              execution time.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="gpusum"><a name="gpusum" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#gpusum" name="gpusum" shape="rect">gpusum[:base] -- GPU Summary (kernels + memory operations)</a></h3>
                        <div class="body">
                           <p class="p">Arguments</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">base</strong> - Optional argument, if given, will cause summary to
                                 be over the base name of the kernel, rather than the templated name.
                              </li>
                           </ul>
                           <p class="p">Output: All time values given in nanoseconds</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">Time(%)</strong> : Percentage of <strong class="ph b">Total Time</strong></li>
                              <li class="li"><strong class="ph b">Total Time</strong> : The total time used by all executions of this
                                 kernel
                              </li>
                              <li class="li"><strong class="ph b">Instances</strong> : The number of executions of this object
                              </li>
                              <li class="li"><strong class="ph b">Average</strong> : The average execution time of this kernel
                              </li>
                              <li class="li"><strong class="ph b">Minimum</strong> : The smallest execution time of this kernel
                              </li>
                              <li class="li"><strong class="ph b">Maximum</strong> : The largest execution time of this kernel
                              </li>
                              <li class="li"><strong class="ph b">Category</strong> : The category of the operation
                              </li>
                              <li class="li"><strong class="ph b">Name</strong> : The name of the kernel
                              </li>
                           </ul>
                           <p class="p">
                              This report provides a summary of CUDA kernels and memory operations,
                              and their execution times. Note that the <strong class="ph b">Time(%)</strong> column is
                              calculated using a summation of the <strong class="ph b">Total Time</strong> column, and
                              represents that kernel's or memory operation's percent of the
                              execution time of the kernels and memory operations listed, and not a\
                              percentage of the application wall or CPU execution time.
                           </p>
                           <p class="p">
                              This report combines data from the <samp class="ph codeph">gpukernsum</samp> and
                              <samp class="ph codeph">gpumemtimesum</samp> reports. This report is very similar to
                              output of the command <samp class="ph codeph">nvprof --print-gpu-summary</samp>.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="gputrace"><a name="gputrace" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#gputrace" name="gputrace" shape="rect">gputrace -- CUDA GPU Trace</a></h3>
                        <div class="body">
                           <p class="p">Arguments - None</p>
                           <p class="p">Output:</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">Start</strong> : Start time of trace event in seconds
                              </li>
                              <li class="li"><strong class="ph b">Duration</strong> : Length of event in nanoseconds
                              </li>
                              <li class="li"><strong class="ph b">CorrId</strong> : Correlation ID
                              </li>
                              <li class="li"><strong class="ph b">GrdX, GrdY, GrdZ</strong> : Grid values
                              </li>
                              <li class="li"><strong class="ph b">BlkX, BlkY, BlkZ</strong> : Block values
                              </li>
                              <li class="li"><strong class="ph b">Reg/Trd</strong> : Registers per thread
                              </li>
                              <li class="li"><strong class="ph b">StcSMem</strong> : Size of Static Shared Memory
                              </li>
                              <li class="li"><strong class="ph b">DymSMem</strong> : Size of Dynamic Shared Memory
                              </li>
                              <li class="li"><strong class="ph b">Bytes</strong> : Size of memory operation
                              </li>
                              <li class="li"><strong class="ph b">Thru</strong> : Throughput in MB per Second
                              </li>
                              <li class="li"><strong class="ph b">SrcMemKd</strong> : Memcpy source memory kind or memset memory kind
                              </li>
                              <li class="li"><strong class="ph b">DstMemKd</strong> : Memcpy destination memory kind
                              </li>
                              <li class="li"><strong class="ph b">Device</strong> : GPU device name and ID
                              </li>
                              <li class="li"><strong class="ph b">Ctx</strong> : Context ID
                              </li>
                              <li class="li"><strong class="ph b">Strm</strong> : Stream ID
                              </li>
                              <li class="li"><strong class="ph b">Name</strong> : Trace event name
                              </li>
                           </ul>
                           <p class="p">
                              This report displays a trace of CUDA kernels and memory operations.
                              Items are sorted by start time.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="nvtxppsum"><a name="nvtxppsum" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#nvtxppsum" name="nvtxppsum" shape="rect">nvtxppsum -- NVTX Push/Pop Range Summary</a></h3>
                        <div class="body">
                           <p class="p">Arguments - None</p>
                           <p class="p">Output: All time values given in nanoseconds</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">Time(%)</strong> : Percentage of <strong class="ph b">Total Time</strong></li>
                              <li class="li"><strong class="ph b">Total Time</strong> : The total time used by all instances of this range
                              </li>
                              <li class="li"><strong class="ph b">Instances</strong> : The number of instances of this range
                              </li>
                              <li class="li"><strong class="ph b">Average</strong> : The average execution time of this range
                              </li>
                              <li class="li"><strong class="ph b"> Minimum</strong> : The smallest execution time of this range
                              </li>
                              <li class="li"><strong class="ph b">Maximum</strong> : The largest execution time of this range
                              </li>
                              <li class="li"><strong class="ph b">Range</strong> : The name of the range
                              </li>
                           </ul>
                           <p class="p">
                              This report provides a summary of NV Tools Extensions Push/Pop Ranges
                              and their execution times. Note that the <strong class="ph b">Time(%)</strong> column is
                              calculated using a summation of the <strong class="ph b">Total Time</strong> column, and
                              represents that range's percent of the execution time of the ranges
                              listed, and not a percentage of the application wall or CPU execution
                              time.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="openmpevtsum"><a name="openmpevtsum" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#openmpevtsum" name="openmpevtsum" shape="rect">openmpevtsum -- OpenMP Event Summary</a></h3>
                        <div class="body">
                           <p class="p">Arguments - None</p>
                           <p class="p">Output: All time values given in nanoseconds</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">Time(%)</strong> : Percentage of <strong class="ph b">Total Time</strong></li>
                              <li class="li"><strong class="ph b">Total Time</strong> : The total time used by all executions of event type
                              </li>
                              <li class="li"><strong class="ph b">Count</strong> : The number of event type
                              </li>
                              <li class="li"><strong class="ph b">Average</strong> : The average execution time of event type
                              </li>
                              <li class="li"><strong class="ph b">Minimum</strong> : The smallest execution time of event type
                              </li>
                              <li class="li"><strong class="ph b">Maximum</strong> : The largest execution time of event type
                              </li>
                              <li class="li"><strong class="ph b">Name</strong> : The name of the event
                              </li>
                           </ul>
                           <p class="p">
                              This report provides a summary of OpenMP events and their execution
                              times. Note that the <strong class="ph b">Time(%)</strong> column is calculated using a
                              summation of the <strong class="ph b">Total Time</strong> column, and represents that event
                              type's percent of the execution time of the events listed, and not a
                              percentage of the application wall or CPU execution time.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="osrtsum"><a name="osrtsum" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#osrtsum" name="osrtsum" shape="rect">osrtsum -- OS Runtime Summary</a></h3>
                        <div class="body">
                           <p class="p">Arguments - None</p>
                           <p class="p">Output: All time values given in nanoseconds</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">Time(%)</strong> : Percentage of <strong class="ph b">Total Time</strong></li>
                              <li class="li"><strong class="ph b">Total Time</strong> : The total time used by all executions of this function
                              </li>
                              <li class="li"><strong class="ph b">Num Calls</strong> : The number of calls to this function
                              </li>
                              <li class="li"><strong class="ph b">Average</strong> : The average execution time of this function
                              </li>
                              <li class="li"><strong class="ph b">Minimum</strong> : The smallest execution time of this function
                              </li>
                              <li class="li"><strong class="ph b">Maximum</strong> : The largest execution time of this function
                              </li>
                              <li class="li"><strong class="ph b">Name</strong> : The name of the function
                              </li>
                           </ul>
                           <p class="p">
                              This report provides a summary of operating system functions and their
                              execution times. Note that the <strong class="ph b">Time(%)</strong> column is calculated
                              using a summation of the <strong class="ph b">Total Time</strong> column, and represents that
                              function's percent of the execution time of the functions listed, and
                              not a percentage of the application wall or CPU execution time.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="vulkanmarkerssum"><a name="vulkanmarkerssum" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#vulkanmarkerssum" name="vulkanmarkerssum" shape="rect">vulkanmarkerssum -- Vulkan Range Summary</a></h3>
                        <div class="body">
                           <p class="p">Arguments - None</p>
                           <p class="p">Output: All time values given in nanoseconds</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">Time(%)</strong> : Percentage of <strong class="ph b">Total Time</strong></li>
                              <li class="li"><strong class="ph b">Total Time</strong> : The total time used by all executions of this function
                              </li>
                              <li class="li"><strong class="ph b">Instances</strong> : The number of instances of this range
                              </li>
                              <li class="li"><strong class="ph b">Average</strong> : The average execution time of this function
                              </li>
                              <li class="li"><strong class="ph b">Minimum</strong> : The smallest execution time of this function
                              </li>
                              <li class="li"><strong class="ph b">Maximum</strong> : The largest execution time of this function
                              </li>
                              <li class="li"><strong class="ph b">StdDev</strong> : The standard deviation of execution time of this range
                              </li>
                              <li class="li"><strong class="ph b">Range</strong> : The name of the range
                              </li>
                           </ul>
                           <p class="p">
                              This report provides a summary of Vulkan debug markers on the CPU,
                              and their execution times. Note that the <strong class="ph b">Time(%)</strong> column is calculated
                              using a summation of the <strong class="ph b">Total Time</strong> column, and represents that
                              function's percent of the execution time of the functions listed, and
                              not a percentage of the application wall or CPU execution time.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="pixsum"><a name="pixsum" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#pixsum" name="pixsum" shape="rect">pixsum -- PIX Range Summary</a></h3>
                        <div class="body">
                           <p class="p">Arguments - None</p>
                           <p class="p">Output: All time values given in nanoseconds</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">Time(%)</strong> : Percentage of <strong class="ph b">Total Time</strong></li>
                              <li class="li"><strong class="ph b">Total Time</strong> : The total time used by all executions of this function
                              </li>
                              <li class="li"><strong class="ph b">Instances</strong> : The number of instances of this range
                              </li>
                              <li class="li"><strong class="ph b">Average</strong> : The average execution time of this function
                              </li>
                              <li class="li"><strong class="ph b">Minimum</strong> : The smallest execution time of this function
                              </li>
                              <li class="li"><strong class="ph b">Maximum</strong> : The largest execution time of this function
                              </li>
                              <li class="li"><strong class="ph b">StdDev</strong> : The standard deviation of execution time of this range
                              </li>
                              <li class="li"><strong class="ph b">Range</strong> : The name of the range
                              </li>
                           </ul>
                           <p class="p">
                              This report provides a summary of PIX CPU debug markers,
                              and their execution times. Note that the <strong class="ph b">Time(%)</strong> column is calculated
                              using a summation of the <strong class="ph b">Total Time</strong> column, and represents that
                              function's percent of the execution time of the functions listed, and
                              not a percentage of the application wall or CPU execution time.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="khrdebugsum"><a name="khrdebugsum" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#khrdebugsum" name="khrdebugsum" shape="rect">khrdebugsum -- OpenGL KHR_debug Range Summary</a></h3>
                        <div class="body">
                           <p class="p">Arguments - None</p>
                           <p class="p">Output: All time values given in nanoseconds</p>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">Time(%)</strong> : Percentage of <strong class="ph b">Total Time</strong></li>
                              <li class="li"><strong class="ph b">Total Time</strong> : The total time used by all executions of this function
                              </li>
                              <li class="li"><strong class="ph b">Instances</strong> : The number of instances of this range
                              </li>
                              <li class="li"><strong class="ph b">Average</strong> : The average execution time of this function
                              </li>
                              <li class="li"><strong class="ph b">Minimum</strong> : The smallest execution time of this function
                              </li>
                              <li class="li"><strong class="ph b">Maximum</strong> : The largest execution time of this function
                              </li>
                              <li class="li"><strong class="ph b">StdDev</strong> : The standard deviation of execution time of this range
                              </li>
                              <li class="li"><strong class="ph b">Range</strong> : The name of the range
                              </li>
                           </ul>
                           <p class="p">
                              This report provides a summary of OpenGL KHR_debug CPU PUSH/POP debug Ranges,
                              and their execution times. Note that the <strong class="ph b">Time(%)</strong> column is calculated
                              using a summation of the <strong class="ph b">Total Time</strong> column, and represents that
                              function's percent of the execution time of the functions listed, and
                              not a percentage of the application wall or CPU execution time.
                           </p>
                        </div>
                     </div>
                  </div>
                  <div class="topic nested1" id="existing-formatters"><a name="existing-formatters" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#existing-formatters" name="existing-formatters" shape="rect">Report Formatters Shipped With Nsight Systems</a></h3>
                     <div class="body">
                        <p class="p">The following formats are available in <span class="ph">Nsight Systems</span></p>
                     </div>
                     <div class="topic nested2" id="column-format"><a name="column-format" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#column-format" name="column-format" shape="rect">Column</a></h3>
                        <div class="body">
                           <p class="p">Usage:</p>
                           <p class="p"><samp class="ph codeph">column[:nohdr][:nolimit][:nofmt][:&lt;width&gt;[:&lt;width&gt;]...]
                                 </samp></p>
                           <p class="p">Arguments</p>
                           <ul class="ul">
                              <li class="li">nohdr : Do not display the header</li>
                              <li class="li">nolimit : Remove 100 character limit from auto-width columns
                                 Note: This can result in extremely wide columns.
                              </li>
                              <li class="li">nofmt : Do not reformat numbers.</li>
                              <li class="li">&lt;width&gt;... : Define the explicit width of one or more
                                 columns. If the value "." is given, the column will auto-adjust.
                                 If a width of 0 is given, the column will not be displayed.
                              </li>
                           </ul>
                           <p class="p">
                              The column formatter presents data in vertical text columns. It is
                              primarily designed to be a human-readable format for displaying data
                              on a console display.
                           </p>
                           <p class="p">
                              Text data will be left-justified, while numeric data will be
                              right-justified. If the data overflows the available column
                              width, it will be marked with a "…" character, to indicate
                              the data values were clipped. Clipping always occurs on the
                              right-hand side, even for numeric data.
                           </p>
                           <p class="p">
                              Numbers will be reformatted to make easier to visually scan and
                              understand. This includes adding thousands-separators. This process
                              requires that the string representation of the number is converted
                              into its native representation (integer or floating point) and then
                              converted back into a string representation to print. This conversion
                              process attempts to preserve elements of number presentation, such as
                              the number of decimal places, or the use of scientific notation, but
                              the conversion is not always perfect (the number should always be the
                              same, but the presentation may not be). To disable the reformatting
                              process, use the argument <samp class="ph codeph">nofmt</samp>.
                           </p>
                           <p class="p">
                              If no explicit width is given, the columns auto-adjust their
                              width based off the header size and the first 100 lines of data.
                              This auto-adjustment is limited to a maximum width of 100
                              characters. To allow larger auto-width columns, pass the initial
                              argument <samp class="ph codeph">nolimit</samp>. If the first 100 lines do not
                              calculate the correct column width, it is suggested that explicit
                              column widths be provided.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="table-formatter"><a name="table-formatter" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#table-formatter" name="table-formatter" shape="rect">Table</a></h3>
                        <div class="body">
                           <p class="p">Usage:</p>
                           <p class="p"><samp class="ph codeph">table[:nohdr][:nolimit][:nofmt][:&lt;width&gt;[:&lt;width&gt;]...]
                                 </samp></p>
                           <p class="p">Arguments</p>
                           <ul class="ul">
                              <li class="li">nohdr : Do not display the header</li>
                              <li class="li">nolimit : Remove 100 character limit from auto-width columns
                                 Note: This can result in extremely wide columns.
                              </li>
                              <li class="li">nofmt : Do not reformat numbers.</li>
                              <li class="li">&lt;width&gt;... : Define the explicit width of one or more
                                 columns. If the value "." is given, the column will auto-adjust.
                                 If a width of 0 is given, the column will not be displayed.
                              </li>
                           </ul>
                           <p class="p">
                              The table formatter presents data in vertical text columns inside
                              text boxes. Other than the lines between columns, it is identical to
                              the <strong class="ph b">column</strong> formatter.
                              
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="csv-formatter"><a name="csv-formatter" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#csv-formatter" name="csv-formatter" shape="rect">CSV</a></h3>
                        <div class="body">
                           <p class="p">Usage:</p>
                           <p class="p"><samp class="ph codeph">csv[:nohdr]</samp></p>
                           <p class="p">Arguments</p>
                           <ul class="ul">
                              <li class="li">nohdr : Do not display the header</li>
                           </ul>
                           <p class="p">
                              The csv formatter outputs data as comma-separated values. This
                              format is commonly used for import into other data applications,
                              such as spread-sheets and databases.
                           </p>
                           <p class="p">
                              There are many different standards for CSV files. Most
                              differences are in how escapes are handled, meaning data values that
                              contain a comma or space.
                           </p>
                           <p class="p">This CSV formatter will escape commas by surrounding the whole
                              value in double-quotes.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="tsv-formatter"><a name="tsv-formatter" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#tsv-formatter" name="tsv-formatter" shape="rect">TSV</a></h3>
                        <div class="body">
                           <p class="p">Usage:</p>
                           <p class="p"><samp class="ph codeph">tsv[:nohdr][:esc]</samp></p>
                           <p class="p">Arguments</p>
                           <ul class="ul">
                              <li class="li">nohdr : Do not display the header</li>
                              <li class="li">esc : escape tab characters, rather than removing them</li>
                           </ul>
                           <p class="p">
                              The tsv formatter outputs data as tab-separated values. This format
                              is sometimes used for import into other data applications, such as
                              spreadsheets and databases.
                           </p>
                           <p class="p">
                              Most TSV import/export systems disallow the tab character in data
                              values. The formatter will normally replace any tab characters with
                              a single space. If the <samp class="ph codeph">esc</samp> argument has been
                              provided, any tab characters will be replaced with the literal
                              characters "\t".
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="json-formatter"><a name="json-formatter" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#json-formatter" name="json-formatter" shape="rect">JSON</a></h3>
                        <div class="body">
                           <p class="p">Usage:</p>
                           <p class="p"><samp class="ph codeph">json</samp></p>
                           <p class="p">Arguments: no arguments</p>
                           <p class="p">
                              The json formatter outputs data as an array of JSON objects. Each
                              object represents one line of data, and uses the column names as
                              field labels. All objects have the same fields. The formatter
                              attempts to recognize numeric values, as well as JSON keywords, and
                              converts them. Empty values are passed as an empty string (and not
                              nil, or as a missing field).
                           </p>
                           <p class="p">
                              At this time the formatter does not escape quotes, so if a data
                              value includes double-quotation marks, it will corrupt the JSON file.
                              
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="hdoc-formatter"><a name="hdoc-formatter" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#hdoc-formatter" name="hdoc-formatter" shape="rect">HDoc</a></h3>
                        <div class="body">
                           <p class="p">Usage:</p>
                           <p class="p"><samp class="ph codeph">hdoc[:title=&lt;title&gt;][:css=&lt;URL&gt;]</samp></p>
                           <p class="p">Arguments:</p>
                           <ul class="ul">
                              <li class="li">title : string for HTML document title</li>
                              <li class="li">css : URL of CSS document to include</li>
                           </ul>
                           <p class="p">
                              The hdoc formatter generates a complete, verifiable (mostly),
                              standalone HTML document. It is designed to be opened in a web
                              browser, or included in a larger document via an &lt;iframe&gt;.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="htable-formatter"><a name="htable-formatter" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#htable-formatter" name="htable-formatter" shape="rect">HTable</a></h3>
                        <div class="body">
                           <p class="p">Usage:</p>
                           <p class="p"><samp class="ph codeph">htable</samp></p>
                           <p class="p">Arguments: no arguments</p>
                           <p class="p">
                              The htable formatter outputs a raw HTML &lt;table&gt; without any of
                              the surrounding HTML document. It is designed to be included into
                              a larger HTML document. Although most web browsers will open and
                              display the document, it is better to use the <strong class="ph b">hdoc</strong> format for
                              this type of use.
                           </p>
                        </div>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="migrate-nvprof"><a name="migrate-nvprof" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#migrate-nvprof" name="migrate-nvprof" shape="rect">5.&nbsp;Migrating from NVIDIA nvprof</a></h2>
                  <div class="body"></div>
                  <div class="topic nested1" id="using-nsys-nvprof"><a name="using-nsys-nvprof" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#using-nsys-nvprof" name="using-nsys-nvprof" shape="rect">Using the Nsight Systems CLI nvprof Command</a></h3>
                     <div class="body">
                        <p class="p">The <samp class="ph codeph">nvprof</samp> command of the <span class="ph">Nsight Systems</span>
                           CLI is intended to help former nvprof users transition to
                           <span class="ph">nsys</span>. Many nvprof switches are not supported by
                           <span class="ph">nsys</span>, often because they are now part of NVIDIA
                           Nsight Compute.
                        </p>
                        <p class="p">The full nvprof documentation can be found at
                           <a class="xref" href="https://docs.nvidia.com/cuda/profiler-users-guide" target="_blank" shape="rect">https://docs.nvidia.com/cuda/profiler-users-guide</a>.
                           
                        </p>
                        <p class="p">The nvprof transition guide for Nsight Compute can be found at
                           <a class="xref" href="https://docs.nvidia.com/nsight-compute/NsightComputeCli/index.html#nvprof-guide" target="_blank" shape="rect">
                              https://docs.nvidia.com/nsight-compute/NsightComputeCli/index.html#nvprof-guide</a>.
                           
                        </p>
                        <p class="p">Any nvprof switch not listed below is not supported by the
                           <samp class="ph codeph"><span class="ph">nsys</span> nvprof</samp> command. No additional
                           <span class="ph">nsys</span> functionality is available through this
                           command. New features will not be added to this command in the future.
                           
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="nvprof-switches"><a name="nvprof-switches" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#nvprof-switches" name="nvprof-switches" shape="rect">CLI nvprof Command Switch Options</a></h3>
                     <div class="body">
                        <p class="p">After choosing the <samp class="ph codeph">nvprof</samp> command switch, the following
                           options are available. When you are ready to move to using
                           <span class="ph">Nsight Systems</span> CLI directly, see
                           <a class="xref" href="index.html#cli-profiling" shape="rect">Command Line Options</a>
                           documentation for the <span class="ph">nsys</span> switch(es) given below.
                           Note that the <span class="ph">nsys</span> implementation and output may
                           vary from nvprof.
                        </p>
                        <p class="p"><strong class="ph b">Usage</strong>.
                        </p><pre xml:space="preserve"><span class="ph">nsys</span> nvprof [options]</pre><div class="p">
                           <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                              <tr class="sthead">
                                 <th valign="bottom" align="left" id="d54e9579" class="stentry" width="13.333333333333334%" rowspan="1" colspan="1">Switch</th>
                                 <th valign="bottom" align="left" id="d54e9582" class="stentry" width="13.333333333333334%" rowspan="1" colspan="1">Parameters (Default in Bold)</th>
                                 <th valign="bottom" align="left" id="d54e9585" class="stentry" width="20%" rowspan="1" colspan="1"><span class="ph">nsys</span> switch
                                 </th>
                                 <th valign="bottom" align="left" id="d54e9590" class="stentry" width="53.333333333333336%" rowspan="1" colspan="1">Switch Description</th>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e9579" class="stentry" rowspan="1" colspan="1">--annotate-mpi</td>
                                 <td valign="top" headers="d54e9582" class="stentry" rowspan="1" colspan="1"><strong class="ph b">off</strong>, openmpi, mpich
                                 </td>
                                 <td valign="top" headers="d54e9585" class="stentry" rowspan="1" colspan="1">--trace=mpi AND --mpi-impl</td>
                                 <td valign="top" headers="d54e9590" class="stentry" rowspan="1" colspan="1">Automatically annotate MPI calls with NVTX markers. Specify
                                    the MPI implementation installed on your machine. Only OpenMPI and
                                    MPICH implementations are supported.
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e9579" class="stentry" rowspan="1" colspan="1">--cpu-thread-tracing</td>
                                 <td valign="top" headers="d54e9582" class="stentry" rowspan="1" colspan="1">on, <strong class="ph b">off</strong></td>
                                 <td valign="top" headers="d54e9585" class="stentry" rowspan="1" colspan="1">--trace=osrt</td>
                                 <td valign="top" headers="d54e9590" class="stentry" rowspan="1" colspan="1">Collect information about CPU thread API activity.</td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e9579" class="stentry" rowspan="1" colspan="1">--profile-api-trace</td>
                                 <td valign="top" headers="d54e9582" class="stentry" rowspan="1" colspan="1">none, runtime, driver,<strong class="ph b">all</strong></td>
                                 <td valign="top" headers="d54e9585" class="stentry" rowspan="1" colspan="1">--trace=cuda</td>
                                 <td valign="top" headers="d54e9590" class="stentry" rowspan="1" colspan="1">Turn on/off CUDA runtime and driver API tracing. For
                                    <span class="ph">Nsight Systems</span> there is no separate CUDA runtime and
                                    CUDA driver trace, so selecting <samp class="ph codeph">runtime</samp> or
                                    <samp class="ph codeph">driver</samp> is equivalent to selecting <samp class="ph codeph">all
                                       </samp>.
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e9579" class="stentry" rowspan="1" colspan="1">--profile-from-start</td>
                                 <td valign="top" headers="d54e9582" class="stentry" rowspan="1" colspan="1"><strong class="ph b">on</strong>, off
                                 </td>
                                 <td valign="top" headers="d54e9585" class="stentry" rowspan="1" colspan="1">if off use --capture-range=cudaProfilerApi</td>
                                 <td valign="top" headers="d54e9590" class="stentry" rowspan="1" colspan="1">Enable/disable profiling from the start of the application.
                                    If disabled, the application can use {cu,cuda}Profiler{Start,Stop}
                                    to turn on/off profiling.
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e9579" class="stentry" rowspan="1" colspan="1">-t,--timeout</td>
                                 <td valign="top" headers="d54e9582" class="stentry" rowspan="1" colspan="1">&lt;nanoseconds&gt; default=<strong class="ph b">0</strong></td>
                                 <td valign="top" headers="d54e9585" class="stentry" rowspan="1" colspan="1">--duration=seconds</td>
                                 <td valign="top" headers="d54e9590" class="stentry" rowspan="1" colspan="1">If greater than 0, stop the collection and kill the launched
                                    application after timeout seconds. nvprof started counting when the
                                    CUDA driver is initialized. <span class="ph">nsys</span> starts counting
                                    immediately.
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e9579" class="stentry" rowspan="1" colspan="1">--cpu-profiling</td>
                                 <td valign="top" headers="d54e9582" class="stentry" rowspan="1" colspan="1">on, <strong class="ph b">off</strong></td>
                                 <td valign="top" headers="d54e9585" class="stentry" rowspan="1" colspan="1">--sampling=cpu</td>
                                 <td valign="top" headers="d54e9590" class="stentry" rowspan="1" colspan="1">Turn on/off CPU profiling</td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e9579" class="stentry" rowspan="1" colspan="1">--openacc-profiling</td>
                                 <td valign="top" headers="d54e9582" class="stentry" rowspan="1" colspan="1"><strong class="ph b">on</strong>, off
                                 </td>
                                 <td valign="top" headers="d54e9585" class="stentry" rowspan="1" colspan="1">--trace=openacc to turn on</td>
                                 <td valign="top" headers="d54e9590" class="stentry" rowspan="1" colspan="1">Enable/disable recording information from the OpenACC
                                    profiling interface. Note: OpenACC profiling interface depends on the
                                    presence of the OpenACC runtime. For supported runtimes, see <strong class="ph b">CUDA
                                       Trace</strong> section of documentation
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e9579" class="stentry" rowspan="1" colspan="1">-o, --export-profile</td>
                                 <td valign="top" headers="d54e9582" class="stentry" rowspan="1" colspan="1">&lt;filename&gt;</td>
                                 <td valign="top" headers="d54e9585" class="stentry" rowspan="1" colspan="1">--output={filename} and/or --export=sqlite</td>
                                 <td valign="top" headers="d54e9590" class="stentry" rowspan="1" colspan="1">Export named file to be imported or opened in the
                                    <span class="ph">Nsight Systems</span> GUI. %q{ENV_VAR} in string will be
                                    replaced with the set value of the environment variable. If not set
                                    this is an error. %h in the string is replaced with the system
                                    hostname. %% in the string is replaced with %. %p in the string is not
                                    supported currently. Any other character following % is illegal. The
                                    default is report1, with the number incrementing to avoid overwriting
                                    files, in users working directory.
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e9579" class="stentry" rowspan="1" colspan="1">-f, --force-overwrite</td>
                                 <td valign="top" headers="d54e9582" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                 <td valign="top" headers="d54e9585" class="stentry" rowspan="1" colspan="1">--force-overwrite=true</td>
                                 <td valign="top" headers="d54e9590" class="stentry" rowspan="1" colspan="1">Force overwriting all output files with same name.</td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e9579" class="stentry" rowspan="1" colspan="1">-h, --help</td>
                                 <td valign="top" headers="d54e9582" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                 <td valign="top" headers="d54e9585" class="stentry" rowspan="1" colspan="1">--help</td>
                                 <td valign="top" headers="d54e9590" class="stentry" rowspan="1" colspan="1">Print <span class="ph">Nsight Systems</span> CLI help
                                 </td>
                              </tr>
                              <tr class="strow">
                                 <td valign="top" headers="d54e9579" class="stentry" rowspan="1" colspan="1">-V, --version</td>
                                 <td valign="top" headers="d54e9582" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                                 <td valign="top" headers="d54e9585" class="stentry" rowspan="1" colspan="1">--version</td>
                                 <td valign="top" headers="d54e9590" class="stentry" rowspan="1" colspan="1">Print <span class="ph">Nsight Systems</span> CLI version information
                                    
                                 </td>
                              </tr>
                           </table>
                        </div>
                     </div>
                  </div>
                  <div class="topic nested1" id="nvprof-next-steps"><a name="nvprof-next-steps" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#nvprof-next-steps" name="nvprof-next-steps" shape="rect">Next Steps</a></h3>
                     <div class="body">
                        <p class="p">NVIDIA Visual Profiler (NVVP) and NVIDIA nvprof are deprecated. New GPUs
                           and features will not be supported by those tools. We encourage you to
                           make the move to <span class="ph">Nsight Systems</span> now. For additional
                           information, suggestions, and rationale, see the blog series in
                           <a class="xref" href="../additional-info/index.html#other-resources" shape="rect">Other Resources</a>.
                           
                        </p>
                        <p class="p"></p>
                        <p class="p hr"></p>
                        <p class="p"></p>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="docker-profiling"><a name="docker-profiling" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#docker-profiling" name="docker-profiling" shape="rect">6.&nbsp;Profiling in a Docker on Linux Devices</a></h2>
                  <div class="body">
                     <p class="p"><strong class="ph b">Collecting data within a Docker</strong></p>
                     <p class="p">
                        The following information assumes the reader is knowledgeable regarding
                        Docker containers. For further information about Docker use in general, see
                        the <a class="xref" href="https://docs.docker.com" target="_blank" shape="rect">
                           Docker documentation</a>.
                     </p>
                     <p class="p"><strong class="ph b">Enable Docker Collection</strong></p>
                     <p class="p">When starting the Docker to perform a <span class="ph">Nsight Systems</span>
                        collection, additional steps are required to enable the
                        <samp class="ph codeph">perf_event_open</samp> system call. This is required in order to
                        utilize the Linux kernel’s perf subsystem which provides sampling
                        information to <span class="ph">Nsight Systems</span>.
                     </p>
                     <p class="p">There are three ways to enable the <samp class="ph codeph">perf_event_open</samp> syscall.
                        You can enable it by using the <samp class="ph codeph">--privileged=true</samp> switch,
                        adding <samp class="ph codeph">--cap-add=SYS_ADMIN</samp> switch to your docker run
                        command file, or you can enable it by setting the seccomp security profile
                        if your system meets the requirements.
                     </p>
                     <p class="p">Secure computing mode (seccomp) is a feature of the Linux kernel that can be
                        used to restrict an application's access. This feature is available only if
                        the kernel is enabled with seccomp support. To check for seccomp support:
                     </p><pre xml:space="preserve">$ grep CONFIG_SECCOMP= /boot/config-$(uname -r)</pre><p class="p">The official Docker documentation says:</p><pre xml:space="preserve">"Seccomp profiles require seccomp 2.2.1 which is not available on Ubuntu 14.04, Debian Wheezy, or Debian Jessie. To use seccomp on these distributions, you must download the latest static Linux binaries (rather than packages)." </pre><div class="p">
                        Download the default seccomp profile file, default.json, relevant to your
                        Docker version. If <samp class="ph codeph">perf_event_open</samp> is already listed in the
                        file as guarded by <samp class="ph codeph">CAP_SYS_ADMIN</samp>, then remove the
                        <samp class="ph codeph">perf_event_open</samp> line. Add the following lines under
                        "syscalls" and save the resulting file as
                        <samp class="ph codeph">default_with_perf.json</samp>.
                        <pre xml:space="preserve">{
    "name": "perf_event_open",
    "action": "SCMP_ACT_ALLOW",
    "args": []
},</pre></div>
                     <p class="p">
                        Then you will be able to use the following switch when starting the
                        Docker to apply the new seccomp profile.
                     </p><pre xml:space="preserve">--security-opt seccomp=default_with_perf.json</pre><p class="p"><strong class="ph b">Launch Docker Collection</strong></p>
                     <p class="p">Here is an example command that has been used to launch a Docker for testing
                        with <span class="ph">Nsight Systems</span>:
                     </p><pre xml:space="preserve">sudo nvidia-docker run --network=host --security-opt seccomp=default_with_perf.json --rm -ti caffe-demo2 bash</pre><p class="p">There is a known issue where Docker collections terminate prematurely with
                        older versions of the driver and the CUDA Toolkit. If collection is ending
                        unexpectedly, please update to the latest versions.
                     </p>
                     <p class="p">After the Docker has been started, use the <span class="ph">Nsight Systems</span> CLI
                        to launch a collection within the Docker. The resulting .qdstrm file can be
                        imported into the <span class="ph">Nsight Systems</span> host like any other CLI
                        result.
                     </p>
                  </div>
                  <div class="topic nested1" id="gui_vnc_container"><a name="gui_vnc_container" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#gui_vnc_container" name="gui_vnc_container" shape="rect">GUI VNC container</a></h3>
                     <div class="body">
                        <p class="p"><span class="ph">Nsight Systems</span> provides a build script to build a self
                           isolated Docker container with the <span class="ph">Nsight Systems</span> GUI and VNC
                           server.
                        </p>
                        <p class="p">You can find the build.py script in the <samp class="ph codeph">host-linux-x64/Scripts/VncContainer</samp>
                           directory (or similar on other architectures) under your <span class="ph">Nsight Systems</span> installation directory.
                           You will need to have <a class="xref" href="https://www.docker.com" target="_blank" shape="rect">Docker</a>,
                           and Python 3.5 or later.
                        </p>
                        <p class="p"><strong class="ph b">Available Parameters</strong></p>
                        <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                           <tr class="sthead">
                              <th valign="bottom" align="left" id="d54e9991" class="stentry" width="12.5%" rowspan="1" colspan="1">Short Name</th>
                              <th valign="bottom" align="left" id="d54e9994" class="stentry" width="25%" rowspan="1" colspan="1">Full Name</th>
                              <th valign="bottom" align="left" id="d54e9997" class="stentry" width="62.5%" rowspan="1" colspan="1">Description</th>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e9991" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                              <td valign="top" headers="d54e9994" class="stentry" rowspan="1" colspan="1">--vnc-password</td>
                              <td valign="top" headers="d54e9997" class="stentry" rowspan="1" colspan="1">(optional) Default password for VNC access (at least 6
                                 characters). If it is specified and empty - will be asked during the
                                 build. Can be changed when running a container.
                              </td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e9991" class="stentry" rowspan="1" colspan="1">-aba</td>
                              <td valign="top" headers="d54e9994" class="stentry" rowspan="1" colspan="1">--additional-build-arguments</td>
                              <td valign="top" headers="d54e9997" class="stentry" rowspan="1" colspan="1">(optional) Additional arguments, which will be passed to the
                                 "docker build" command.
                              </td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e9991" class="stentry" rowspan="1" colspan="1">-hd</td>
                              <td valign="top" headers="d54e9994" class="stentry" rowspan="1" colspan="1">--nsys-host-directory</td>
                              <td valign="top" headers="d54e9997" class="stentry" rowspan="1" colspan="1">(optional) The directory with <span class="ph">Nsight Systems</span>
                                 host binaries (with GUI).
                              </td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e9991" class="stentry" rowspan="1" colspan="1">-td</td>
                              <td valign="top" headers="d54e9994" class="stentry" rowspan="1" colspan="1">--nsys-target-directory</td>
                              <td valign="top" headers="d54e9997" class="stentry" rowspan="1" colspan="1">(optional, repeatable) The directory with
                                 <span class="ph">Nsight Systems</span> target binaries (can be specified multiple
                                 times).
                              </td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e9991" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                              <td valign="top" headers="d54e9994" class="stentry" rowspan="1" colspan="1">--tigervnc</td>
                              <td valign="top" headers="d54e9997" class="stentry" rowspan="1" colspan="1">(optional) Use TigerVNC instead of x11vnc.</td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e9991" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                              <td valign="top" headers="d54e9994" class="stentry" rowspan="1" colspan="1">--http</td>
                              <td valign="top" headers="d54e9997" class="stentry" rowspan="1" colspan="1">(optional) Install noVNC in the Docker container for HTTP
                                 access.
                              </td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e9991" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                              <td valign="top" headers="d54e9994" class="stentry" rowspan="1" colspan="1">--rdp</td>
                              <td valign="top" headers="d54e9997" class="stentry" rowspan="1" colspan="1">(optional) Install xRDP in the Docker for RDP access.</td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e9991" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                              <td valign="top" headers="d54e9994" class="stentry" rowspan="1" colspan="1">--geometry</td>
                              <td valign="top" headers="d54e9997" class="stentry" rowspan="1" colspan="1">(optional) Default VNC server resolution in the format
                                 WidthxHeight (default 1920x1080).
                              </td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e9991" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                              <td valign="top" headers="d54e9994" class="stentry" rowspan="1" colspan="1">--build-directory</td>
                              <td valign="top" headers="d54e9997" class="stentry" rowspan="1" colspan="1">(optional) The directory to save temporary files (with the
                                 write access for the current user). By default, script or tmp directory
                                 will be used.
                              </td>
                           </tr>
                        </table>
                        <p class="p"><strong class="ph b">Ports</strong></p>
                        <p class="p">These ports can be published from the container to provide access
                           to the Docker container:
                        </p>
                        <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                           <tr class="sthead">
                              <th valign="bottom" align="left" id="d54e10123" class="stentry" width="30%" rowspan="1" colspan="1">Port</th>
                              <th valign="bottom" align="left" id="d54e10126" class="stentry" width="30%" rowspan="1" colspan="1">Purpose</th>
                              <th valign="bottom" align="left" id="d54e10129" class="stentry" width="40%" rowspan="1" colspan="1">Condition</th>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e10123" class="stentry" rowspan="1" colspan="1">TCP 5900</td>
                              <td valign="top" headers="d54e10126" class="stentry" rowspan="1" colspan="1">Port for VNC access</td>
                              <td valign="top" headers="d54e10129" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e10123" class="stentry" rowspan="1" colspan="1">TCP 80 (optional)</td>
                              <td valign="top" headers="d54e10126" class="stentry" rowspan="1" colspan="1">Port for HTTP access to noVNC server</td>
                              <td valign="top" headers="d54e10129" class="stentry" rowspan="1" colspan="1">Container is build with "--http" parameter</td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e10123" class="stentry" rowspan="1" colspan="1">TCP 3389 (optional)</td>
                              <td valign="top" headers="d54e10126" class="stentry" rowspan="1" colspan="1">Port for RDP access</td>
                              <td valign="top" headers="d54e10129" class="stentry" rowspan="1" colspan="1">Container is build with "--rdp" parameter</td>
                           </tr>
                        </table>
                        <p class="p"><strong class="ph b">Volumes</strong></p>
                        <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                           <tr class="sthead">
                              <th valign="bottom" align="left" id="d54e10177" class="stentry" width="12.5%" rowspan="1" colspan="1">Docker folder</th>
                              <th valign="bottom" align="left" id="d54e10180" class="stentry" width="25%" rowspan="1" colspan="1">Purpose</th>
                              <th valign="bottom" align="left" id="d54e10183" class="stentry" width="62.5%" rowspan="1" colspan="1">Description</th>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e10177" class="stentry" rowspan="1" colspan="1">/mnt/host</td>
                              <td valign="top" headers="d54e10180" class="stentry" rowspan="1" colspan="1">Root path for shared folders</td>
                              <td valign="top" headers="d54e10183" class="stentry" rowspan="1" colspan="1">Folder owned by the Docker user (inner content can be
                                 accessed from <span class="ph">Nsight Systems</span> GUI)
                              </td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e10177" class="stentry" rowspan="1" colspan="1">/mnt/host/Projects</td>
                              <td valign="top" headers="d54e10180" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                              <td valign="top" headers="d54e10183" class="stentry" rowspan="1" colspan="1">Folder with projects and reports, created by
                                 <span class="ph">Nsight Systems</span> UI in container
                              </td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e10177" class="stentry" rowspan="1" colspan="1">/mnt/host/logs</td>
                              <td valign="top" headers="d54e10180" class="stentry" rowspan="1" colspan="1">Folder with inner services logs</td>
                              <td valign="top" headers="d54e10183" class="stentry" rowspan="1" colspan="1">May be useful to send reports to developers</td>
                           </tr>
                        </table>
                        <p class="p"><strong class="ph b">Environment variables</strong></p>
                        <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                           <tr class="sthead">
                              <th valign="bottom" align="left" id="d54e10237" class="stentry" width="50%" rowspan="1" colspan="1">Variable Name</th>
                              <th valign="bottom" align="left" id="d54e10240" class="stentry" width="50%" rowspan="1" colspan="1">Purpose</th>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e10237" class="stentry" rowspan="1" colspan="1">VNC_PASSWORD</td>
                              <td valign="top" headers="d54e10240" class="stentry" rowspan="1" colspan="1">Password for VNC access (at least 6 characters)</td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e10237" class="stentry" rowspan="1" colspan="1">NSYS_WINDOW_WIDTH</td>
                              <td valign="top" headers="d54e10240" class="stentry" rowspan="1" colspan="1">Width of VNC server display (in pixels)</td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e10237" class="stentry" rowspan="1" colspan="1">NSYS_WINDOW_HEIGHT</td>
                              <td valign="top" headers="d54e10240" class="stentry" rowspan="1" colspan="1">Height of VNC server display (in pixels)</td>
                           </tr>
                        </table>
                        <p class="p"><strong class="ph b">Examples</strong></p>
                        <p class="p">With VNC access on port 5916:</p><pre xml:space="preserve">sudo docker run -p 5916:5900/tcp -ti nsys-ui-vnc:1.0</pre><p class="p">With VNC access on port 5916 and HTTP access on port 8080:</p><pre xml:space="preserve">sudo docker run -p 5916:5900/tcp -p 8080:80/tcp -ti nsys-ui-vnc:1.0</pre><p class="p">With VNC access on port 5916, HTTP access on port 8080 and RDP access on
                           port 33890:
                        </p><pre xml:space="preserve">sudo docker run -p 5916:5900/tcp -p 8080:80/tcp -p 33890:3389/tcp -ti nsys-ui-vnc:1.0</pre><p class="p">With VNC access on port 5916, shared "HOME" folder from the host, VNC server
                           resolution 3840x2160, and custom VNC password
                        </p><pre xml:space="preserve">sudo docker run -p 5916:5900/tcp -v $HOME:/mnt/host/home -e NSYS_WINDOW_WIDTH=3840 -e NSYS_WINDOW_HEIGHT=2160 -e VNC_PASSWORD=7654321 -ti nsys-ui-vnc:1.0</pre><p class="p">With VNC access on port 5916, shared "HOME" folder from the host, and the
                           projects folder to access reports created by <span class="ph">Nsight Systems</span>
                           GUI in container
                        </p><pre xml:space="preserve">sudo docker run -p 5916:5900/tcp -v $HOME:/mnt/host/home -v /opt/NsysProjects:/mnt/host/Projects -ti nsys-ui-vnc:1.0</pre></div>
                  </div>
               </div>
               <div class="topic nested0" id="direct3d"><a name="direct3d" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#direct3d" name="direct3d" shape="rect">7.&nbsp;Direct3D Trace</a></h2>
                  <div class="body">
                     <p class="p"><span class="ph">Nsight Systems</span> has the ability to trace both the Direct3D 11 API
                        and the Direct3D 12 API on Windows targets.
                     </p>
                  </div>
                  <div class="topic nested1" id="d3d11-trace"><a name="d3d11-trace" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#d3d11-trace" name="d3d11-trace" shape="rect">7.1.&nbsp;D3D11 API trace</a></h3>
                     <div class="body">
                        <p class="p"><span class="ph">Nsight Systems</span> can capture information about Direct3D 11
                           API calls made by the profiled process. This includes capturing the
                           execution time of D3D11 API functions, performance markers, and frame
                           durations.
                        </p>
                        <blockquote class="lq">
                           <p class="p"><img class="image" src="graphics/DX11_API_Row.png" height="-1" width="-1" alt="D3D11 API Trace"></img></p>
                        </blockquote>
                        <p class="p"><strong class="ph b">SLI Trace</strong></p>
                        <p class="p">Trace SLI queries and peer-to-peer transfers of D3D11
                           applications. Requires SLI hardware and an active SLI profile definition
                           in the NVIDIA console.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="d3d12-trace"><a name="d3d12-trace" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#d3d12-trace" name="d3d12-trace" shape="rect">7.2.&nbsp;D3D12 API Trace</a></h3>
                     <div class="body">
                        <p class="p">Direct3D 12 is a low-overhead 3D graphics and compute API for Microsoft
                           Windows. Information about Direct3D 12 can be found at the <a class="xref" href="https://docs.microsoft.com/en-us/windows/desktop/direct3d12/directx-12-programming-guide" target="_blank" shape="rect">Direct3D 12 Programming Guide</a>.
                        </p>
                        <p class="p"><span class="ph">Nsight Systems</span> can capture information about Direct3D 12
                           usage by the profiled process. This includes capturing the execution time
                           of D3D12 API functions, corresponding workloads executed on the GPU,
                           performance markers, and frame durations.
                        </p>
                        <p class="p"><img class="image" src="graphics/dx12_overview.png" height="-1" width="-1" alt="D3D12 overview picture"></img></p>
                        <p class="p">The Command List Creation row displays time periods when command lists
                           were being created. This enables developers to improve their application’s
                           multi-threaded command list creation. Command list creation time period is
                           measured between the call to <samp class="ph codeph">ID3D12GraphicsCommandList::Reset</samp>
                           and the call to <samp class="ph codeph">ID3D12GraphicsCommandList::Close</samp>.
                        </p>
                        <p class="p"><img class="image" src="graphics/dx12_commandlist_creation.png" height="-1" width="-1" alt="D3D12 commandlist creation"></img></p>
                        <p class="p">The GPU row shows a compressed view of the D3D12 queue activity, color-coded
                           by the queue type. Expanding it will show the individual queues and their corresponding API calls.
                        </p><br clear="none"></br><img class="image" src="graphics/dx12_gpu_aggregated.png" alt="D3D12 GPU aggregated"></img><br clear="none"></br><p class="p">A Command Queue row is displayed for each D3D12 command queue created by
                           the profiled application. The row’s header displays the queue's running
                           index and its type (Direct, Compute, Copy).
                        </p>
                        <p class="p"><img class="image" src="graphics/dx12_commandqueues.png" height="-1" width="-1" alt="D3D12 command queue overview"></img></p>
                        <p class="p">The DX12 API Memory Ops row displays all API memory operations and
                           non-persistent resource mappings. Event ranges in the row are color-coded
                           by the heap type they belong to (Default, Readback, Upload, Custom, or
                           CPU-Visible VRAM), with usage warnings highlighted in yellow. A breakdown
                           of the operations can be found by expanding the row to show rows for each
                           individual heap type.
                        </p>
                        <p class="p">The following operations and warnings are shown:</p>
                        <ul class="ul">
                           <li class="li">
                              <p class="p">Calls to <samp class="ph codeph">ID3D12Device::CreateCommittedResource</samp>,
                                 <samp class="ph codeph">ID3D12Device4::CreateCommittedResource1</samp>, and
                                 <samp class="ph codeph">ID3D12Device8::CreateCommittedResource2</samp></p>
                              <ul class="ul">
                                 <li class="li">
                                    <p class="p">A warning will be reported if
                                       <samp class="ph codeph">D3D12_HEAP_FLAG_CREATE_NOT_ZEROED</samp> is not set in the
                                       method's <samp class="ph codeph">HeapFlags</samp> parameter
                                    </p>
                                 </li>
                              </ul>
                           </li>
                           <li class="li">
                              <p class="p">Calls to <samp class="ph codeph">ID3D12Device::CreateHeap</samp> and
                                 <samp class="ph codeph">ID3D12Device4::CreateHeap1</samp></p>
                              <ul class="ul">
                                 <li class="li">
                                    <p class="p">A warning will be reported if
                                       <samp class="ph codeph">D3D12_HEAP_FLAG_CREATE_NOT_ZEROED</samp> is not set in the
                                       <samp class="ph codeph">Flags</samp> field of the method's <samp class="ph codeph">pDesc</samp>
                                       parameter
                                    </p>
                                 </li>
                              </ul>
                           </li>
                           <li class="li">
                              <p class="p">Calls to <samp class="ph codeph">ID3D12Resource::ReadFromSubResource</samp></p>
                              <ul class="ul">
                                 <li class="li">
                                    <p class="p">A warning will be reported if the read is to a
                                       <samp class="ph codeph">D3D12_CPU_PAGE_PROPERTY_WRITE_COMBINE</samp> CPU page or from a
                                       <samp class="ph codeph">D3D12_HEAP_TYPE_UPLOAD</samp> resource
                                    </p>
                                 </li>
                              </ul>
                           </li>
                           <li class="li">
                              <p class="p">Calls to <samp class="ph codeph">ID3D12Resource::WriteToSubResource</samp></p>
                              <ul class="ul">
                                 <li class="li">
                                    <p class="p">A warning will be reported if the write is from a
                                       <samp class="ph codeph">D3D12_CPU_PAGE_PROPERTY_WRITE_BACK</samp> CPU page or to a
                                       <samp class="ph codeph">D3D12_HEAP_TYPE_READBACK</samp> resource
                                    </p>
                                 </li>
                              </ul>
                           </li>
                           <li class="li">
                              <p class="p">Calls to <samp class="ph codeph">ID3D12Resource::Map</samp> and
                                 <samp class="ph codeph">ID3D12Resource::Unmap</samp> will be matched into [Map, Unmap]
                                 ranges for non-persistent mappings. If a mapping range is nested, only
                                 the most external range (reference count = 1) will be shown.
                              </p>
                           </li>
                        </ul>
                        <p class="p"><img class="image" src="graphics/dx12_memory_operations.png" height="-1" width="-1" alt="D3D12 memory operations and usage warning"></img></p>
                        <p class="p">The API row displays time periods where
                           <samp class="ph codeph">ID3D12CommandQueue::ExecuteCommandLists</samp> was called. The
                           GPU Workload row displays time periods where workloads were executed by
                           the GPU. The workload’s type (Graphics, Compute, Copy, etc.) is displayed
                           on the bar representing the workload’s GPU execution.
                        </p>
                        <p class="p"><img class="image" src="graphics/dx12_api_and_related_workload.png" height="-1" width="-1" alt="D3D12 API and Workload"></img></p>
                        <p class="p">In addition, you can see the PIX command queue CPU-side performance
                           markers, GPU-side performance markers and the GPU Command List
                           performance markers, each in their row.
                        </p>
                        <p class="p"><img class="image" src="graphics/dx12_cpu_marker.png" height="-1" width="-1" alt="D3D12 CPU marker"></img></p>
                        <p class="p"><img class="image" src="graphics/dx12_gpu_marker.png" height="-1" width="-1" alt="D3D12 GPU marker"></img></p>
                        <p class="p"><img class="image" src="graphics/dx12_commandlist_marker.png" height="-1" width="-1" alt="D3D12 commandlist marker"></img></p>
                        <p class="p">
                           Clicking on a GPU workload highlights the corresponding
                           <samp class="ph codeph">ID3D12CommandQueue::ExecuteCommandLists</samp>,
                           <samp class="ph codeph">ID3D12GraphicsCommandList::Reset</samp> and
                           <samp class="ph codeph">ID3D12GraphicsCommandList::Close API</samp> calls, and vice
                           versa.
                        </p>
                        <p class="p"><img class="image" src="graphics/dx12_correlation.png" height="-1" width="-1" alt="D3D12 correlation"></img></p>
                        <div class="p">Detecting which CPU thread was blocked by a fence can be difficult in
                           complex apps that run tens of CPU threads. The timeline view displays the
                           3 operations involved:
                           
                           <ul class="ul">
                              <li class="li">
                                 <p class="p">The CPU thread pushing a signal command and fence value into the command
                                    queue. This is displayed on the DX12 Synchronization sub-row of the
                                    calling thread.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">The GPU executing that command, setting the fence value and signaling
                                    the fence. This is displayed on the GPU Queue Synchronization sub-row.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">The CPU thread calling a Win32 wait API to block-wait until the fence is
                                    signaled. This is displayed on the Thread's OS runtime libraries row.
                                 </p>
                              </li>
                           </ul>
                        </div>
                        <p class="p">Clicking one of these will highlight it and the corresponding other
                           two calls.
                        </p>
                        <p class="p"><img class="image" src="graphics/d3d12_fence_synchronization.png" height="-1" width="-1" alt="D3D12 fence sync"></img></p>
                        <p class="p"></p>
                        <p class="p hr"></p>
                        <p class="p"></p>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="wddm-queues"><a name="wddm-queues" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#wddm-queues" name="wddm-queues" shape="rect">8.&nbsp;WDDM Queues</a></h2>
                  <div class="body">
                     <p class="p">
                        The Windows Display Driver Model (WDDM) architecture uses queues to send
                        work packets from the CPU to the GPU. Each D3D device in each process is
                        associated with one or more contexts. Graphics, compute, and copy commands
                        that the profiled application uses are associated with a context, batched in
                        a command buffer, and pushed into the relevant queue associated with that
                        context.
                     </p>
                     <p class="p"><span class="ph">Nsight Systems</span> can capture the state of these queues during
                        the trace session.
                     </p>
                     <p class="p">
                        Enabling the "Collect additional range of ETW events" option will also
                        capture extended DxgKrnl events from the <samp class="ph codeph">Microsoft-Windows-DxgKrnl</samp>
                        provider, such as context status, allocations, sync wait, signal events, etc.
                     </p>
                     <blockquote class="lq">
                        <p class="p"><img class="image" src="graphics/WDDM_Queues.png" alt="WDDM Queues"></img></p>
                     </blockquote>
                     <div class="p">A command buffer in a WDDM queues may have one the following types:
                        
                        <ul class="ul">
                           <li class="li">
                              <p class="p">Render</p>
                           </li>
                           <li class="li">
                              <p class="p">Deferred</p>
                           </li>
                           <li class="li">
                              <p class="p">System</p>
                           </li>
                           <li class="li">
                              <p class="p">MMIOFlip</p>
                           </li>
                           <li class="li">
                              <p class="p">Wait</p>
                           </li>
                           <li class="li">
                              <p class="p">Signal</p>
                           </li>
                           <li class="li">
                              <p class="p">Device</p>
                           </li>
                           <li class="li">
                              <p class="p">Software</p>
                           </li>
                        </ul>
                     </div>
                     <p class="p">It may also be marked as a Present buffer, indicating that the application
                        has finished rendering and requests to display the source surface.
                     </p>
                     <p class="p">See the Microsoft documentation for the WDDM architecture and the
                        DXGKETW_QUEUE_PACKET_TYPE enumeration.
                     </p>
                     <p class="p"></p>
                     <p class="p hr"></p>
                     <p class="p"></p>
                     <p class="p">To retain the .etl trace files captured, so that they can be viewed in
                        other tools (e.g. GPUView), change the "Save ETW log files in project
                        folder" option under "Profile Behavior" in <span class="ph">Nsight Systems</span>'s
                        global Options dialog. The .etl files will appear in the same folder as the
                        .nsys-rep file, accessible by right-clicking the report in the Project Explorer
                        and choosing "Show in Folder...". Data collected from each ETW provider
                        will appear in its own .etl file, and an additional .etl file named
                        "Report XX-Merged-*.etl", containing the events from all captured sources,
                        will be created as well.
                        
                     </p>
                  </div>
               </div>
               <div class="topic nested0" id="wddm-hw-scheduler"><a name="wddm-hw-scheduler" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#wddm-hw-scheduler" name="wddm-hw-scheduler" shape="rect">9.&nbsp;WDDM HW Scheduler</a></h2>
                  <div class="body">
                     <p class="p">
                        When GPU Hardware Scheduling is enabled in Windows 10 or newer version, the
                        Windows Display Driver Model (WDDM) uses the DxgKrnl ETW provider to expose
                        report of NVIDIA GPUs' hardware scheduling context switches.
                     </p>
                     <p class="p"><span class="ph">Nsight Systems</span> can capture these context switch events, and display
                        under the GPUs in the timeline rows titled WDDM HW Scheduler - [HW Queue type].
                        The ranges under each queue will show the process name and PID assoicated with
                        the GPU work during the time period.
                     </p>
                     <p class="p">
                        The events will be captured if GPU Hardware Scheduling is enabled in the Windows
                        System Display settings, and "Collect WDDM Trace" is enabled in the <span class="ph">Nsight Systems</span>
                        Project Settings.
                     </p>
                     <blockquote class="lq">
                        <p class="p"><img class="image" src="graphics/wddm-hw-scheduler.png" alt="WDDM HW Scheduler row for 3D HW Queue"></img></p>
                     </blockquote>
                  </div>
               </div>
               <div class="topic nested0" id="vulkan"><a name="vulkan" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#vulkan" name="vulkan" shape="rect">10.&nbsp;Vulkan API Trace</a></h2>
                  <div class="body">
                     <p class="p"></p>
                  </div>
                  <div class="topic nested1" id="vulkan-overview"><a name="vulkan-overview" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#vulkan-overview" name="vulkan-overview" shape="rect">10.1.&nbsp;Vulkan Overview</a></h3>
                     <div class="body">
                        <p class="p">Vulkan is a low-overhead, cross-platform 3D graphics and compute API,
                           targeting a wide variety of devices from PCs to mobile phones and
                           embedded platforms. The Vulkan API is defined by the Khronos Group.
                           Information about Vulkan and the Khronos Group can be found at the
                           <a class="xref" href="https://www.khronos.org/vulkan/" target="_blank" shape="rect">Khronos Vulkan Site</a>.
                        </p>
                        <p class="p"><span class="ph">Nsight Systems</span> can capture information about Vulkan usage
                           by the profiled process. This includes capturing the execution time of
                           Vulkan API functions, corresponding GPU workloads, debug util labels, and
                           frame durations. Vulkan profiling is supported on both Windows and x86
                           Linux operating systems.
                        </p>
                        <p class="p"><img class="image" src="graphics/vulkan_overview.png" alt="Vulkan overview picture"></img></p>
                        <p class="p">The Command Buffer Creation row displays time periods when command
                           buffers were being created. This enables developers to improve their
                           application’s multi-threaded command buffer creation. Command buffer
                           creation time period is measured between the call to
                           <samp class="ph codeph">vkBeginCommandBuffer</samp> and the call to
                           <samp class="ph codeph">vkEndCommandBuffer</samp>.
                        </p>
                        <p class="p"><img class="image" src="graphics/vulkan_command_buffer_creation.png" alt="Vulkan command buffer creation"></img></p>
                        <p class="p">A Queue row is displayed for each Vulkan queue created by the profiled
                           application. The API sub-row displays time periods where
                           <samp class="ph codeph">vkQueueSubmit</samp> was called. The GPU Workload sub-row
                           displays time periods where workloads were executed by the GPU.
                        </p>
                        <p class="p"><img class="image" src="graphics/vulkan_api_and_related_workload.png" alt="Vulkan API and Workload"></img></p>
                        <p class="p">In addition, you can see <a class="xref" href="https://github.com/KhronosGroup/Vulkan-Docs/blob/master/appendices/VK_EXT_debug_utils.txt" target="_blank" shape="rect">Vulkan debug util labels</a> on both
                           the CPU and the GPU.
                        </p>
                        <p class="p"><img class="image" src="graphics/vulkan_cpu_marker.png" alt="Vulkan CPU marker"></img></p>
                        <p class="p">Clicking on a GPU workload highlights the corresponding
                           <samp class="ph codeph">vkQueueSubmit</samp> call, and vice versa.
                        </p>
                        <p class="p"><img class="image" src="graphics/vulkan_correlation.png" alt="Vulkan correlation"></img></p>
                        <p class="p">The Vulkan Memory Operations row contains an aggregation of all the Vulkan host-side memory
                           operations, such as host-blocking writes and reads or non-persistent map-unmap ranges.
                        </p>
                        <p class="p">The row is separated into sub-rows by heap index and memory type - the tooltip for each
                           row and the ranges inside show the heap flags and the memory property flags.
                        </p>
                        <p class="p"><img class="image" src="graphics/vulkan_memory_operations.png" alt="Vulkan Memory Operations"></img></p>
                        <p class="p"><img class="image" src="graphics/vulkan_memory_operations_row.png" alt="Vulkan Memory Operations"></img></p>
                        <p class="p"></p>
                     </div>
                  </div>
                  <div class="topic nested1" id="vulkan-pipeline-creation"><a name="vulkan-pipeline-creation" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#vulkan-pipeline-creation" name="vulkan-pipeline-creation" shape="rect">10.2.&nbsp;Pipeline Creation Feedback</a></h3>
                     <div class="body">
                        <p class="p">
                           When tracing target application calls to Vulkan pipeline creation APIs,
                           <span class="ph">Nsight Systems</span> leverages the Pipeline Creation Feedback
                           extension to collect more details about the duration of individual
                           pipeline creation stages.
                        </p>
                        <p class="p">
                           See <a class="xref" href="https://www.khronos.org/registry/vulkan/specs/1.1-extensions/html/vkspec.html#VkPipelineCreationFeedbackEXT" target="_blank" shape="rect">Pipeline Creation Feedback extension</a>
                           for details about this extension.
                        </p>
                        <p class="p">
                           Vulkan pipeline creation feedback is available on NVIDIA driver release
                           435 or later.
                        </p>
                        <p class="p"><img class="image" src="graphics/vulkan-pipeline-creation-feedback.png" alt="Vulkan pipeline creation feedback"></img></p>
                     </div>
                  </div>
                  <div class="topic nested1" id="vulkan-gpu-trace"><a name="vulkan-gpu-trace" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#vulkan-gpu-trace" name="vulkan-gpu-trace" shape="rect">10.3.&nbsp;Vulkan GPU Trace Notes</a></h3>
                     <div class="body">
                        <div class="p">
                           <ul class="ul">
                              <li class="li">
                                 <p class="p">Vulkan GPU trace is available only when tracing apps that use NVIDIA
                                    GPUs.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">The endings of Vulkan Command Buffers execution ranges on Compute and
                                    Transfer queues may appear earlier on the timeline than their actual
                                    occurrence.
                                 </p>
                              </li>
                           </ul>
                        </div>
                        <p class="p"></p>
                        <p class="p hr"></p>
                        <p class="p"></p>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="stutter-analysis"><a name="stutter-analysis" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#stutter-analysis" name="stutter-analysis" shape="rect">11.&nbsp;Stutter Analysis</a></h2>
                  <div class="body">
                     <div class="section" id="stutter-analysis__stutter-overview"><a name="stutter-analysis__stutter-overview" shape="rect">
                           <!-- --></a><h2 class="title sectiontitle">Stutter Analysis Overview</h2>
                        <p class="p"><span class="ph">Nsight Systems</span> on Windows targets displays stutter
                           analysis visualization aids for profiled graphics applications that use
                           either OpenGL, D3D11, D3D12 or Vulkan, as detailed below in the
                           following sections.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="fps-overview"><a name="fps-overview" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#fps-overview" name="fps-overview" shape="rect">11.1.&nbsp;FPS Overview</a></h3>
                     <div class="body">
                        <p class="p">The Frame Duration section displays frame durations on both the CPU and
                           the GPU.
                        </p>
                        <blockquote class="lq">
                           <p class="p"><img class="image" src="graphics/fps_overview.png" alt="FPS overview"></img></p>
                        </blockquote>
                        <div class="p">The frame duration row displays live FPS statistics for the current
                           timeline viewport. Values shown are:
                           
                           <ol class="ol">
                              <li class="li">
                                 <p class="p">Number of CPU frames shown of the total number captured</p>
                              </li>
                              <li class="li">
                                 <p class="p">Average, minimal, and maximal CPU frame time of the currently
                                    displayed time range
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">Average FPS value for the currently displayed frames</p>
                              </li>
                              <li class="li">
                                 <p class="p">The 99th percentile value of the frame lengths (such that only 1%
                                    of the frames in the range are longer than this value).
                                 </p>
                              </li>
                           </ol>
                           
                           The values will update automatically when scrolling, zooming or filtering
                           the timeline view.
                           
                        </div>
                        <blockquote class="lq">
                           <p class="p"><img class="image" src="graphics/stutter_row.png" alt="FPS stutter row"></img></p>
                        </blockquote>
                        <p class="p">The stutter row highlights frames that are significantly longer than the
                           other frames in their immediate vicinity.
                        </p>
                        <p class="p">The stutter row uses an algorithm that compares the duration of each
                           frame to the median duration of the surrounding 19 frames. Duration
                           difference under 4 milliseconds is never considered a stutter, to avoid
                           cluttering the display with frames whose absolute stutter is small and
                           not noticeable to the user.
                        </p>
                        <div class="p">For example, if the stutter threshold is set at 20%:
                           
                           <ol class="ol">
                              <li class="li">
                                 <p class="p">Median duration is 10 ms. Frame with 13 ms time will not be reported
                                    (relative difference &gt; 20%, absolute difference &lt; 4 ms)
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">Median duration is 60 ms. Frame with 71 ms time will not be reported
                                    (relative difference &lt; 20%, absolute difference &gt; 4 ms)
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">Median duration is 60 ms. Frame with 80 ms is a stutter (relative
                                    difference &gt; 20%, absolute difference &gt; 4 ms, both conditions met)
                                 </p>
                              </li>
                           </ol>
                        </div>
                        <p class="p"><strong class="ph b">OSC detection</strong></p>
                        <p class="p">The "19 frame window median" algorithm by itself may not work well with
                           some cases of "oscillation" (consecutive fast and slow frames), resulting
                           in some false positives. The median duration is not meaningful in cases
                           of oscillation and can be misleading.
                        </p>
                        <div class="p">To address the issue and identify if oscillating frames, the following
                           method is applied:
                           
                           <ol class="ol">
                              <li class="li">
                                 <p class="p">For every frame, calculate the median duration, 1st and 3rd quartiles of
                                    19-frames window.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">Calculate the delta and ratio between 1st and 3rd quartiles.</p>
                              </li>
                              <li class="li">
                                 <p class="p">If the 90th percentile of 3rd – 1st quartile delta array &gt; 4 ms AND
                                    the 90th percentile of 3rd/1st quartile array &gt; 1.2 (120%) then mark
                                    the results with "OSC" text.
                                 </p>
                              </li>
                           </ol>
                        </div>
                        <p class="p">Right-clicking the Frame Duration row caption lets you choose the target
                           frame rate (30, 60, 90 or custom frames per second).
                        </p>
                        <p class="p"><img class="image" src="graphics/fps_pick.png" alt="FPS pick"></img></p>
                        <p class="p">By clicking the Customize FPS Display option, a customization dialog pops
                           up. In the dialog, you can now define the frame duration threshold to
                           customize the view of the potentially problematic frames. In addition,
                           you can define the threshold for the stutter analysis frames.
                        </p>
                        <p class="p"><img class="image" src="graphics/fps_customizations.png" alt="fps_customizations"></img></p>
                        <div class="p">Frame duration bars are color coded:
                           
                           <ul class="ul">
                              <li class="li">
                                 <p class="p">Green, the frame duration is shorter than required by the target FPS
                                    ratio.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">Yellow, duration is slightly longer than required by the target FPS rate.</p>
                              </li>
                              <li class="li">
                                 <p class="p">Red, duration far exceeds that required to maintain the target FPS rate.</p>
                              </li>
                           </ul>
                        </div>
                        <p class="p"><img class="image" src="graphics/bad_fps.png" alt="Bad FPS"></img></p>
                        <div class="p">The CPU Frame Duration row displays the CPU frame duration measured
                           between the ends of consecutive frame boundary calls:
                           
                           <ul class="ul">
                              <li class="li">
                                 <p class="p">The OpenGL frame boundaries are
                                    <samp class="ph codeph">eglSwapBuffers/glXSwapBuffers/SwapBuffers</samp> calls.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">The D3D11 and D3D12 frame boundaries are
                                    <samp class="ph codeph">IDXGISwapChainX::Present</samp> calls.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">The Vulkan frame boundaries are <samp class="ph codeph">vkQueuePresentKHR</samp>
                                    calls.
                                 </p>
                              </li>
                           </ul>
                        </div>
                        <p class="p">The timing of the actual calls to the frame boundary calls can be seen
                           in the blue bar at the bottom of the CPU frame duration row
                        </p>
                        <div class="p">The GPU Frame Duration row displays the time measured between
                           
                           <ul class="ul">
                              <li class="li">
                                 <p class="p">The start time of the first GPU workload execution of this frame.</p>
                              </li>
                              <li class="li">
                                 <p class="p">The start time of the first GPU workload execution of the next frame.</p>
                              </li>
                           </ul>
                        </div>
                        <p class="p"><strong class="ph b">Reflex SDK</strong></p>
                        <p class="p">NVIDIA Reflex SDK is a series of NVAPI calls that allow applications to integrate the
                           Ultra Low Latency driver feature more directly into their game to further optimize
                           synchronization between simulation and rendering stages and lower the latency between
                           user input and final image rendering.
                           For more details about Reflex SDK, see <a class="xref" href="https://developer.nvidia.com/reflex/" target="_blank" shape="rect">Reflex SDK Site</a>.
                        </p>
                        <p class="p"><span class="ph">Nsight Systems</span> will automatically capture NVAPI functions
                           when either Direct3D 11, Direct3D 12, or Vulkan API trace are enabled.
                        </p>
                        <div class="p">The Reflex SDK row displays timeline ranges for the following types of latency markers:
                           
                           <ul class="ul">
                              <li class="li">
                                 <p class="p">RenderSubmit.</p>
                              </li>
                              <li class="li">
                                 <p class="p">Simulation.</p>
                              </li>
                              <li class="li">
                                 <p class="p">Present.</p>
                              </li>
                              <li class="li">
                                 <p class="p">Driver.</p>
                              </li>
                              <li class="li">
                                 <p class="p">OS Render Queue.</p>
                              </li>
                              <li class="li">
                                 <p class="p">GPU Render.</p>
                              </li>
                           </ul>
                        </div>
                        <p class="p"><img class="image" src="graphics/reflex_sdk.png" alt="Reflex SDK"></img></p>
                        <p class="p"><strong class="ph b">Performance Warnings row</strong></p>
                        <div class="p">This row shows performance warnings and common pitfalls that are automatically detected
                           based on the enabled capture types.
                           Warnings are reported for:
                           
                           <ul class="ul">
                              <li class="li">
                                 <p class="p">ETW performance warnings</p>
                              </li>
                              <li class="li">
                                 <p class="p">Vulkan calls to <samp class="ph codeph">vkQueueSubmit</samp> and D3D12 calls to
                                    <samp class="ph codeph">ID3D12CommandQueue::ExecuteCommandList</samp> that take a longer time to
                                    execute than the total time of the GPU workloads they generated
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><a class="xref" href="topics/d3d12-trace" shape="rect">D3D12 Memory Operation warnings</a></p>
                              </li>
                              <li class="li">
                                 <p class="p">Usage of Vulkan API functions that may adversely affect performance</p>
                              </li>
                              <li class="li">
                                 <p class="p">Creation of a Vulkan device with memory zeroing, whether by physical device default or manually</p>
                              </li>
                              <li class="li">
                                 <p class="p">Vulkan command buffer barrier which can be combined or removed, such as subsequent barriers or read-to-read barriers</p>
                              </li>
                           </ul>
                        </div>
                        <p class="p"><img class="image" src="graphics/fps_performance_warnings.png" alt="Performance Warnings row"></img></p>
                     </div>
                  </div>
                  <div class="topic nested1" id="frame-health"><a name="frame-health" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#frame-health" name="frame-health" shape="rect">11.2.&nbsp;Frame Health</a></h3>
                     <div class="body">
                        <p class="p">
                           The Frame Health row displays actions that took significantly a longer
                           time during the current frame, compared to the median time of the same
                           actions executed during the surrounding 19-frames. This is a great tool
                           for detecting the reason for frame time stuttering. Such actions may be:
                           shader compilation, present, memory mapping, and more.
                           <span class="ph">Nsight Systems</span> measures the accumulated time of such
                           actions in each frame. For example: calculating the accumulated time of
                           shader compilations in each frame and comparing it to the accumulated
                           time of shader compilations in the surrounding 19 frames.
                        </p>
                        <p class="p">Example of a Vulkan frame health row:</p>
                        <p class="p"><img class="image" src="graphics/vulkan_frame_health.png" alt="Frame Health Vulkan"></img></p>
                        <p class="p"><img class="image" src="graphics/dx12_frame_health.png" alt="Frame Health DX12"></img></p>
                     </div>
                  </div>
                  <div class="topic nested1" id="gpu-memory-utilization"><a name="gpu-memory-utilization" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#gpu-memory-utilization" name="gpu-memory-utilization" shape="rect">11.3.&nbsp;GPU Memory Utilization</a></h3>
                     <div class="body">
                        <p class="p">
                           The Memory Utilization row displays the amount of used local GPU memory
                           and the commit limit for each GPU.
                        </p><img class="image" src="graphics/memory_utilization.png" alt="Memory Utilization"></img><p class="p">Note that this is not the same as the CUDA kernel memory allocation
                           graph, see <a class="xref" href="topics/cuda-gpu-memory-graph" shape="rect">CUDA GPU
                              Memory Graph</a> for that functionality.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="vertical-sync"><a name="vertical-sync" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#vertical-sync" name="vertical-sync" shape="rect">11.4.&nbsp;Vertical Synchronization</a></h3>
                     <div class="body">
                        <p class="p">
                           The VSYNC rows display when the monitor's vertical synchronizations
                           occur.
                        </p>
                        <p class="p"><img class="image" src="graphics/vsync.png" alt="Vertical Synchronization"></img></p>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="openmp-trace"><a name="openmp-trace" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#openmp-trace" name="openmp-trace" shape="rect">12.&nbsp;OpenMP Trace</a></h2>
                  <div class="body">
                     <p class="p"><span class="ph">Nsight Systems</span> for Linux x86_64 and Power targets is capable
                        of capturing information about OpenMP events. This functionality is built on
                        the OpenMP Tools Interface (OMPT), full support is available only for runtime
                        libraries supporting tools interface defined in OpenMP 5.0 or greater.
                     </p>
                     <p class="p">As an example, LLVM OpenMP runtime library partially implements tools interface.
                        If you use PGI compiler &lt;= 20.4 to build your OpenMP applications, add -mp=libomp
                        switch to use LLVM OpenMP runtime and enable OMPT based tracing. If you use Clang,
                        make sure the LLVM OpenMP runtime library you link to was compiled with tools
                        interface enabled.
                        
                     </p>
                     <p class="p"><img class="image" src="graphics/select-openmp-trace.png" alt="OpenMP trace selection"></img></p>
                     <p class="p">Only a subset of the OMPT callbacks are processed:
                        
                     </p><pre xml:space="preserve">ompt_callback_parallel_begin
ompt_callback_parallel_end
ompt_callback_sync_region
ompt_callback_task_create
ompt_callback_task_schedule
ompt_callback_implicit_task
ompt_callback_master
ompt_callback_reduction
ompt_callback_task_create
ompt_callback_cancel
ompt_callback_mutex_acquire, ompt_callback_mutex_acquired
ompt_callback_mutex_acquired, ompt_callback_mutex_released
ompt_callback_mutex_released
ompt_callback_work
ompt_callback_dispatch
ompt_callback_flush</pre><blockquote class="lq">
                        <div class="p">
                           <div class="tablenoborder">
                              <table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
                                 <tbody class="tbody">
                                    <tr class="row">
                                       <td class="entry nvidiagreen" align="center" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                          <p class="p">&nbsp;&nbsp;<strong class="ph b">Note:</strong>&nbsp;&nbsp;
                                          </p>
                                       </td>
                                       <td class="entry notebox" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                          <p class="p">The raw OMPT events are used to generate ranges
                                             indicating the runtime of OpenMP operations and
                                             constructs.&nbsp;&nbsp;
                                          </p>
                                       </td>
                                    </tr>
                                 </tbody>
                              </table>
                           </div>
                        </div>
                     </blockquote>
                     <p class="p">Example screenshot:</p>
                     <p class="p"><img class="image" src="graphics/openmp-trace.png" alt="OpenMP API trace"></img></p>
                     <p class="p"></p>
                     <p class="p hr"></p>
                     <p class="p"></p>
                  </div>
               </div>
               <div class="topic nested0" id="os-runtime-trace"><a name="os-runtime-trace" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#os-runtime-trace" name="os-runtime-trace" shape="rect">13.&nbsp;OS Runtime Libraries Trace</a></h2>
                  <div class="body">
                     <p class="p">On Linux, OS runtime libraries can be traced to gather information about
                        low-level userspace APIs. This traces the system call wrappers and thread
                        synchronization interfaces exposed by the C runtime and POSIX Threads
                        (pthread) libraries. This does not perform a complete runtime library API
                        trace, but instead focuses on the functions that can take a long time to
                        execute, or could potentially cause your thread be unscheduled from the
                        CPU while waiting for an event to complete. OS runtime trace is not
                        available for Windows targets.
                     </p>
                     <div class="p">OS runtime tracing complements and enhances sampling information by:
                        
                        <ol class="ol">
                           <li class="li">
                              <p class="p">Visualizing when the process is communicating with the hardware, controlling
                                 resources, performing multi-threading synchronization or interacting with
                                 the kernel scheduler.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">Adding additional thread states by correlating how OS runtime libraries
                                 traces affect the thread scheduling:
                              </p>
                              <ul class="ul">
                                 <li class="li">
                                    <p class="p"><strong class="ph b">Waiting</strong> — the thread is not scheduled on a CPU, it is inside of an
                                       OS runtime libraries trace and is believed to be waiting on the firmware
                                       to complete a request.
                                    </p>
                                 </li>
                                 <li class="li">
                                    <p class="p"><strong class="ph b">In OS runtime library function</strong> — the thread is scheduled on a CPU
                                       and inside of an OS runtime libraries trace. If the trace represents a
                                       system call, the process is likely running in kernel mode.
                                    </p>
                                 </li>
                              </ul>
                           </li>
                           <li class="li">
                              <p class="p">Collecting backtraces for long OS runtime libraries call. This provides a
                                 way to gather blocked-state backtraces, allowing you to gain more context
                                 about why the thread was blocked so long, yet avoiding unnecessary
                                 overhead for short events.
                              </p>
                           </li>
                        </ol>
                     </div>
                     <blockquote class="lq">
                        <p class="p"><img class="image" src="graphics/os-runtime-row-x86.png" alt="OS runtime libraries row"></img></p>
                     </blockquote>
                     <p class="p">To enable OS runtime libraries tracing from <span class="ph">Nsight Systems</span>:
                     </p>
                     <p class="p"><strong class="ph b">CLI</strong> — Use the <samp class="ph codeph">-t</samp>, <samp class="ph codeph">--trace</samp> option
                        with the <samp class="ph codeph">osrt</samp> parameter. See <a class="xref" href="topics/cli-profiling" shape="rect">
                           Command Line Options</a> for more information.
                     </p>
                     <p class="p"><strong class="ph b">GUI</strong> — Select the <strong class="ph b">Collect OS runtime libraries trace</strong> checkbox.
                     </p>
                     <blockquote class="lq">
                        <p class="p"><img class="image" src="graphics/configure-os-runtime-trace.png" alt="Configure OS runtime libraries trace"></img></p>
                     </blockquote>
                     <p class="p">You can also use <strong class="ph b">Skip if shorter than</strong>. This will skip calls shorter
                        than the given threshold. Enabling this option will improve performances as
                        well as reduce noise on the timeline. We strongly encourage you to skip OS
                        runtime libraries call shorter than 1 μs.
                     </p>
                  </div>
                  <div class="topic nested1" id="osrt-locking-resource"><a name="osrt-locking-resource" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#osrt-locking-resource" name="osrt-locking-resource" shape="rect">13.1.&nbsp;Locking a Resource</a></h3>
                     <div class="body">
                        <p class="p">The functions listed below receive a special treatment. If the tool
                           detects that the resource is already acquired by another thread and will
                           induce a blocking call, we always trace it. Otherwise, it will never be
                           traced.
                        </p><pre xml:space="preserve">pthread_mutex_lock
pthread_rwlock_rdlock
pthread_rwlock_wrlock
pthread_spin_lock
sem_wait</pre><p class="p">Note that even if a call is determined as potentially blocking, there is
                           a chance that it may not actually block after a few cycles have elapsed.
                           The call will still be traced in this scenario.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="osrt-limitations"><a name="osrt-limitations" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#osrt-limitations" name="osrt-limitations" shape="rect">13.2.&nbsp;Limitations</a></h3>
                     <div class="body">
                        <ul class="ul">
                           <li class="li">
                              <p class="p"><span class="ph">Nsight Systems</span> only traces syscall wrappers exposed by the
                                 C runtime. It is not able to trace syscall invoked through assembly code.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">Additional thread states, as well as backtrace collection on long calls,
                                 are only enabled if sampling is turned on.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">It is not possible to configure the depth and duration threshold when
                                 collecting backtraces. Currently, only OS runtime libraries calls longer
                                 than 80 μs will generate a backtrace with a maximum of 24 frames. This
                                 limitation will be removed in a future version of the product.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">It is required to compile your application and libraries with the
                                 <samp class="ph codeph">-funwind-tables</samp> compiler flag in order for
                                 <span class="ph">Nsight Systems</span> to unwind the backtraces correctly.
                              </p>
                           </li>
                        </ul>
                     </div>
                  </div>
                  <div class="topic nested1" id="osrt-filters"><a name="osrt-filters" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#osrt-filters" name="osrt-filters" shape="rect">13.3.&nbsp;OS Runtime Libraries Trace Filters</a></h3>
                     <div class="body">
                        <p class="p">The OS runtime libraries tracing is limited to a select list of
                           functions. It also depends on the version of the C runtime linked to the
                           application.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="osrt-default-function"><a name="osrt-default-function" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#osrt-default-function" name="osrt-default-function" shape="rect">13.4.&nbsp;OS Runtime Default Function List</a></h3>
                     <div class="body">
                        <p class="p"><strong class="ph b">Libc system call wrappers</strong></p><pre xml:space="preserve">accept
accept4
acct
alarm
arch_prctl
bind
bpf
brk
chroot
clock_nanosleep
connect
copy_file_range
creat
creat64
dup
dup2
dup3
epoll_ctl
epoll_pwait
epoll_wait
fallocate
fallocate64
fcntl
fdatasync
flock
fork
fsync
ftruncate
futex
ioctl
ioperm
iopl
kill
killpg
listen
membarrier
mlock
mlock2
mlockall
mmap
mmap64
mount
move_pages
mprotect
mq_notify
mq_open
mq_receive
mq_send
mq_timedreceive
mq_timedsend
mremap
msgctl
msgget
msgrcv
msgsnd
msync
munmap
nanosleep
nfsservctl
open
open64
openat
openat64
pause
pipe
pipe2
pivot_root
poll
ppoll
prctl
pread
pread64
preadv
preadv2
preadv64
process_vm_readv
process_vm_writev
pselect6
ptrace
pwrite
pwrite64
pwritev
pwritev2
pwritev64
read
readv
reboot
recv
recvfrom
recvmmsg
recvmsg
rt_sigaction
rt_sigqueueinfo
rt_sigsuspend
rt_sigtimedwait
sched_yield
seccomp
select
semctl
semget
semop
semtimedop
send
sendfile
sendfile64
sendmmsg
sendmsg
sendto
shmat
shmctl
shmdt
shmget
shutdown
sigaction
sigsuspend
sigtimedwait
socket
socketpair
splice
swapoff
swapon
sync
sync_file_range
syncfs
tee
tgkill
tgsigqueueinfo
tkill
truncate
umount2
unshare
uselib
vfork
vhangup
vmsplice
wait
wait3
wait4
waitid
waitpid
write
writev
_sysctl</pre><p class="p"><strong class="ph b">POSIX Threads</strong></p><pre xml:space="preserve">pthread_barrier_wait
pthread_cancel
pthread_cond_broadcast
pthread_cond_signal
pthread_cond_timedwait
pthread_cond_wait
pthread_create
pthread_join
pthread_kill
pthread_mutex_lock
pthread_mutex_timedlock
pthread_mutex_trylock
pthread_rwlock_rdlock
pthread_rwlock_timedrdlock
pthread_rwlock_timedwrlock
pthread_rwlock_tryrdlock
pthread_rwlock_trywrlock
pthread_rwlock_wrlock
pthread_spin_lock
pthread_spin_trylock
pthread_timedjoin_np
pthread_tryjoin_np
pthread_yield
sem_timedwait
sem_trywait
sem_wait</pre><p class="p"><strong class="ph b">I/O</strong></p><pre xml:space="preserve">aio_fsync
aio_fsync64
aio_suspend
aio_suspend64
fclose
fcloseall
fflush
fflush_unlocked
fgetc
fgetc_unlocked
fgets
fgets_unlocked
fgetwc
fgetwc_unlocked
fgetws
fgetws_unlocked
flockfile
fopen
fopen64
fputc
fputc_unlocked
fputs
fputs_unlocked
fputwc
fputwc_unlocked
fputws
fputws_unlocked
fread
fread_unlocked
freopen
freopen64
ftrylockfile
fwrite
fwrite_unlocked
getc
getc_unlocked
getdelim
getline
getw
getwc
getwc_unlocked
lockf
lockf64
mkfifo
mkfifoat
posix_fallocate
posix_fallocate64
putc
putc_unlocked
putwc
putwc_unlocked</pre><p class="p"><strong class="ph b">Miscellaneous</strong></p><pre xml:space="preserve">forkpty
popen
posix_spawn
posix_spawnp
sigwait
sigwaitinfo
sleep
system
usleep</pre><p class="p"></p>
                        <p class="p hr"></p>
                        <p class="p"></p>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="nvtx-trace"><a name="nvtx-trace" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#nvtx-trace" name="nvtx-trace" shape="rect">14.&nbsp;NVTX Trace</a></h2>
                  <div class="body">
                     <p class="p">The NVIDIA Tools Extension Library (NVTX) is a powerful mechanism that allows
                        users to manually instrument their application. <span class="ph">Nsight Systems</span>
                        can then collect the information and present it on the timeline.
                     </p>
                     <p class="p"><span class="ph">Nsight Systems</span> supports version 3.0 of the NVTX specification.
                        
                     </p>
                     <p class="p">
                        The following features are supported:
                     </p>
                     <ul class="ul">
                        <li class="li">
                           <p class="p">Domains</p><pre xml:space="preserve">nvtxDomainCreate(), nvtxDomainDestroy()</pre><pre xml:space="preserve">nvtxDomainRegisterString()</pre></li>
                        <li class="li">
                           <p class="p">Push-pop ranges (nested ranges that start and end in the same thread).</p><pre xml:space="preserve">nvtxRangePush(), nvtxRangePushEx()</pre><pre xml:space="preserve">nvtxRangePop()</pre><pre xml:space="preserve">nvtxDomainRangePushEx()</pre><pre xml:space="preserve">nvtxDomainRangePop()</pre></li>
                        <li class="li">
                           <p class="p">Start-end ranges (ranges that are global to the process and are not
                              restricted to a single thread)
                           </p><pre xml:space="preserve">nvtxRangeStart(), nvtxRangeStartEx()</pre><pre xml:space="preserve">nvtxRangeEnd()</pre><pre xml:space="preserve">nvtxDomainRangeStartEx()</pre><pre xml:space="preserve">nvtxDomainRangeEnd()</pre></li>
                        <li class="li">
                           <p class="p">Marks</p><pre xml:space="preserve">nvtxMark(), nvtxMarkEx()</pre><pre xml:space="preserve">nvtxDomainMarkEx()</pre></li>
                        <li class="li">
                           <p class="p">Thread names</p><pre xml:space="preserve">nvtxNameOsThread()</pre></li>
                        <li class="li">
                           <p class="p">Categories</p><pre xml:space="preserve">nvtxNameCategory()</pre><pre xml:space="preserve">nvtxDomainNameCategory()</pre></li>
                     </ul>
                     <p class="p">To learn more about specific features of NVTX, please refer to the NVTX
                        header file: <samp class="ph codeph">nvToolsExt.h</samp> or the
                        <a class="xref" href="https://nvidia.github.io/NVTX/" target="_blank" shape="rect">NVTX documentation</a>.
                     </p>
                     <div class="p">To use NVTX in your application, follow these steps:
                        
                        <ol class="ol">
                           <li class="li">
                              <p class="p">Add <samp class="ph codeph">#include "nvtx3/nvToolsExt.h"</samp> in your source code. The
                                 nvtx3 directory is located in the <span class="ph">Nsight Systems</span> package in the
                                 Target-&lt;architecture&gt;/nvtx/include directory and is available via
                                 github at <a class="xref" href="http://github.com/NVIDIA/NVTX" target="_blank" shape="rect">http://github.com/NVIDIA/NVTX</a>.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">Add the following compiler flag: <samp class="ph codeph">-ldl</samp></p>
                           </li>
                           <li class="li">
                              <p class="p">Add calls to the NVTX API functions. For example, try adding
                                 <samp class="ph codeph">nvtxRangePush("main")</samp> in the beginning of the
                                 <samp class="ph codeph">main()</samp> function, and <samp class="ph codeph">nvtxRangePop()</samp> just
                                 before the return statement in the end.
                              </p>
                              <p class="p">For convenience in C++ code, consider adding a wrapper that implements
                                 RAII (resource acquisition is initialization) pattern, which would
                                 guarantee that every range gets closed.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">In the project settings, select the <strong class="ph b">Collect NVTX trace</strong> checkbox.
                              </p>
                           </li>
                        </ol>
                     </div>
                     <p class="p">In addition, by enabling the "Insert NVTX Marker hotkey" option it is
                        possible to add NVTX markers to a running non-console applications by
                        pressing the F11 key. These will appear in the report under the NVTX Domain
                        named "HotKey markers".
                     </p>
                     <p class="p">Typically calls to NVTX functions can be left in the source code even if the
                        application is not being built for profiling purposes, since the overhead is
                        very low when the profiler is not attached.
                     </p>
                     <p class="p">NVTX is not intended to annotate very small pieces of code that are being called
                        very frequently.  A good rule of thumb to use: if code being annotated usually
                        takes less than 1 microsecond to execute, adding an NVTX range around this code
                        should be done carefully.
                     </p>
                     <blockquote class="lq">
                        <div class="p">
                           <div class="tablenoborder">
                              <table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
                                 <tbody class="tbody">
                                    <tr class="row">
                                       <td class="entry nvidiagreen" align="center" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                          <p class="p">&nbsp;&nbsp;<strong class="ph b">Note:</strong>&nbsp;&nbsp;
                                          </p>
                                       </td>
                                       <td class="entry notebox" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                          <p class="p">Range annotations should be matched carefully.  If many ranges are
                                             opened but not closed, <span class="ph">Nsight Systems</span> has no meaningful way to
                                             visualize it.  A rule of thumb is to not have more than a couple dozen ranges
                                             open at any point in time.  <span class="ph">Nsight Systems</span> does not support
                                             reports with many unclosed ranges.
                                          </p>
                                       </td>
                                    </tr>
                                 </tbody>
                              </table>
                           </div>
                        </div>
                     </blockquote>
                     <p class="p"><strong class="ph b">NVTX Domains and Categories</strong></p>
                     <p class="p"></p>
                     <p class="p">NVTX domains enable scoping of annotations. Unless specified
                        differently, all events and annotations are in the default domain.
                        Additionally, categories can be used to group events.
                     </p>
                     <p class="p"><span class="ph">Nsight Systems</span> gives the user the ability to include or
                        exclude NVTX events from a particular domain. This can be especially useful
                        if you are profiling across multiple libraries and are only interested in
                        nvtx events from some of them.
                     </p>
                     <p class="p"><img class="image" src="graphics/nvtx-domain.png" alt="NVTX domain selection screen"></img></p>
                     <p class="p">This functionality is also available from the CLI. See the CLI
                        documentation for <samp class="ph codeph">--nvtx-domain-include</samp> and
                        <samp class="ph codeph">--nvtx-domain-exclude</samp> for more details.
                     </p>
                     <p class="p">Categories that are set in by the user will be recognized and displayed
                        in the GUI.
                     </p>
                     <p class="p"><img class="image" src="graphics/nvtx-screenshot.png" alt="NVTX screenshot with domains and categories"></img></p>
                     <p class="p"></p>
                     <p class="p hr"></p>
                     <p class="p"></p>
                  </div>
               </div>
               <div class="topic nested0" id="cuda-trace"><a name="cuda-trace" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#cuda-trace" name="cuda-trace" shape="rect">15.&nbsp;CUDA Trace</a></h2>
                  <div class="body">
                     <p class="p"><span class="ph">Nsight Systems</span> is capable of capturing information about CUDA
                        execution in the profiled process.
                     </p>
                     <div class="p">The following information can be collected and presented on the timeline in
                        the report:
                        
                        <ul class="ul">
                           <li class="li">
                              <p class="p">CUDA API trace — trace of CUDA Runtime and CUDA Driver calls made by the
                                 application.
                              </p>
                              <ul class="ul">
                                 <li class="li">
                                    <p class="p">CUDA Runtime calls typically start with <samp class="ph codeph">cuda</samp> prefix
                                       (e.g. <samp class="ph codeph">cudaLaunch</samp>).
                                    </p>
                                 </li>
                                 <li class="li">
                                    <p class="p">CUDA Driver calls typically start with <samp class="ph codeph">cu</samp> prefix (e.g.
                                       <samp class="ph codeph">cuDeviceGetCount</samp>).
                                    </p>
                                 </li>
                              </ul>
                           </li>
                           <li class="li">
                              <p class="p">CUDA workload trace — trace of activity happening on the GPU, which
                                 includes memory operations (e.g., Host-to-Device memory copies) and
                                 kernel executions. Within the threads that use the CUDA API, additional
                                 child rows will appear in the timeline tree.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">On <span class="ph">Nsight Systems Workstation Edition</span>, cuDNN and cuBLAS API tracing and
                                 OpenACC tracing.
                              </p>
                           </li>
                        </ul>
                     </div>
                     <p class="p"><img class="image" src="graphics/cuda-thread-rows.png" alt="CUDA thread rows"></img></p>
                     <p class="p">Near the bottom of the timeline row tree, the GPU node will appear and
                        contain a CUDA node.  Within the CUDA node, each CUDA context used within
                        the process will be shown along with its corresponding CUDA streams.  Steams
                        will contain memory operations and kernel launches on the GPU. Kernel
                        launches are represented by blue, while memory transfers are displayed in
                        red.
                     </p>
                     <p class="p"><img class="image" src="graphics/cuda-gpu-rows.png" alt="CUDA GPU rows"></img></p>
                     <p class="p">
                        The easiest way to capture CUDA information is to launch the process from
                        <span class="ph">Nsight Systems</span>, and it will setup the environment for you. To
                        do so, simply set up a normal launch and select the <strong class="ph b">Collect CUDA trace</strong>
                        checkbox.
                     </p>
                     <p class="p">For <span class="ph">Nsight Systems Workstation Edition</span> this looks like:
                     </p>
                     <blockquote class="lq">
                        <p class="p"><img class="image" src="graphics/configure-cuda-trace-x86.png" alt="Configure CUDA trace"></img></p>
                     </blockquote>
                     <p class="p">For <span class="ph">Nsight Systems Embedded Platforms Edition</span> this looks like:
                     </p>
                     <blockquote class="lq">
                        <p class="p"><img class="image" src="graphics/configure-cuda-trace.png" alt="Configure CUDA trace"></img></p>
                     </blockquote>
                     <p class="p">Additional configuration parameters are available:</p>
                     <ul class="ul">
                        <li class="li">
                           <p class="p"><strong class="ph b">Collect backtraces for API calls longer than X seconds</strong> - turns on
                              collection of CUDA API backtraces and sets the minimum time a CUDA API
                              event must take before its backtraces are collected. Setting this value
                              too low can cause high application overhead and seriously increase the
                              size of your results file.
                           </p>
                        </li>
                        <li class="li">
                           <p class="p"><strong class="ph b">Flush data periodically</strong> — specifies the period after which an
                              attempt to flush CUDA trace data will be made.  Normally, in order to
                              collect full CUDA trace, the application needs to finalize the device used
                              for CUDA work (call <samp class="ph codeph">cudaDeviceReset()</samp>, and then let the
                              application gracefully exit (as opposed to crashing).
                           </p>
                           <p class="p">This option allows flushing CUDA trace data even before the device is
                              finalized.  However, it might introduce additional overhead to a random
                              CUDA Driver or CUDA Runtime API call.
                           </p>
                        </li>
                        <li class="li">
                           <p class="p"><strong class="ph b">Skip some API calls</strong> — avoids tracing insignificant CUDA Runtime API
                              calls (namely, <samp class="ph codeph">cudaConfigureCall()</samp>,
                              <samp class="ph codeph">cudaSetupArgument()</samp>,
                              <samp class="ph codeph">cudaHostGetDevicePointers()</samp>).  Not tracing these functions
                              allows <span class="ph">Nsight Systems</span> to significantly reduce the profiling
                              overhead, without losing any interesting data. (See CUDA Trace Filters,
                              below)
                           </p>
                        </li>
                        <li class="li">
                           <p class="p"><strong class="ph b">Collect GPU Memory Usage</strong> - collects information used to
                              generate a graph of CUDA allocated memory across time. Note that this will
                              increase overhead. See section on <strong class="ph b">CUDA GPU Memory Allocation Graph</strong> below.
                              
                           </p>
                        </li>
                        <li class="li">
                           <p class="p"><strong class="ph b">Collect Unified Memory CPU page faults</strong> - collects information on page faults
                              	  that occur when CPU code tries to access a memory page that resides on the device. See
                              	  section on <strong class="ph b">Unified Memory CPU Page Faults</strong> in the <strong class="ph b">Unified Memory Transfer Trace</strong>
                              	  documentation below.
                           </p>
                        </li>
                        <li class="li">
                           <p class="p"><strong class="ph b">Collect Unified Memory GPU page faults</strong> - collects information on page faults
                              	  that occur when GPU code tries to access a memory page that resides on the CPU. See
                              	  section on <strong class="ph b">Unified Memory GPU Page Faults</strong> in the <strong class="ph b">Unified Memory Transfer Trace</strong>
                              	  documentation below.
                           </p>
                        </li>
                        <li class="li">
                           <p class="p"><strong class="ph b">Collect CUDA Graph trace </strong> - by default, CUDA tracing will
                              collect and expose information on a whole graph basis. The user can opt to
                              collect on a node per node basis. See section on <strong class="ph b">CUDA Graph Trace</strong>
                              below.
                           </p>
                        </li>
                        <li class="li">
                           <p class="p">For <span class="ph">Nsight Systems Workstation Edition</span>, <strong class="ph b">Collect cuDNN trace</strong>,
                              <strong class="ph b">Collect cuBLAS trace</strong>, <strong class="ph b">Collect OpenACC trace</strong> - selects which
                              (if any) extra libraries that depend on CUDA to trace.
                           </p>
                           <p class="p">OpenACC versions 2.0, 2.5, and 2.6 are supported when using PGI runtime
                              version 15.7 or greater and not compiling statically. In order to
                              differentiate constructs, a PGI runtime of 16.1 or later is required.
                              Note that <span class="ph">Nsight Systems Workstation Edition</span> does not support the GCC
                              implementation of OpenACC at this time.
                           </p>
                        </li>
                     </ul>
                     <p class="p">Please note that if your application crashes before all collected CUDA
                        trace data has been copied out, some or all data might be lost and not
                        present in the report.
                     </p>
                  </div>
                  <div class="topic nested1" id="cuda-gpu-memory-graph"><a name="cuda-gpu-memory-graph" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#cuda-gpu-memory-graph" name="cuda-gpu-memory-graph" shape="rect">15.1.&nbsp;CUDA GPU Memory Allocation Graph</a></h3>
                     <div class="body">
                        <p class="p">When the <strong class="ph b">Collect GPU Memory Usage</strong> option is selected from the
                           <strong class="ph b">Collect CUDA trace</strong> option set, <span class="ph">Nsight Systems</span> will track
                           CUDA GPU memory allocations and deallocations and present a graph of this
                           information in the timeline. This is not the same as the GPU memory graph
                           generated during stutter analysis on the Windows target (see
                           <a class="xref" href="topics/gpu-memory-utilization" shape="rect">Stutter Memory Trace</a>)
                        </p>
                        <p class="p">Below, in the report on the left, memory is allocated and freed during the
                           collection. In the report on the right, memory is allocated, but not freed
                           during the collection.
                        </p>
                        <p class="p"><img class="image" src="graphics/memory_allocation_graph1.png" alt="CUDA memory allocation graphs where the memory is or is not released during application run"></img></p>
                        <p class="p">Here is another example, where allocations are happening on multiple GPUs</p>
                        <p class="p"><img class="image" src="graphics/memory_allocation_graph2.png" alt="CUDA memory usage graph on multiple threads"></img></p>
                     </div>
                  </div>
                  <div class="topic nested1" id="unified-memory-trace"><a name="unified-memory-trace" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#unified-memory-trace" name="unified-memory-trace" shape="rect">15.2.&nbsp;Unified Memory Transfer Trace</a></h3>
                     <div class="body">
                        <p class="p">
                           For <span class="ph">Nsight Systems Workstation Edition</span>, Unified Memory (also called
                           Managed Memory) transfer trace is enabled automatically in
                           <span class="ph">Nsight Systems</span> when CUDA trace is selected. It incurs no
                           overhead in programs that do not perform any Unified Memory transfers.
                           Data is displayed in the Managed Memory area of the timeline:
                        </p>
                        <p class="p"><img class="image" src="graphics/uvm.png" alt="UVM trace"></img></p>
                        <p class="p"><strong class="ph b">HtoD transfer</strong> indicates the CUDA kernel accessed managed memory
                           that was residing on the host, so the kernel execution paused and
                           transferred the data to the device. Heavy traffic here will incur
                           performance penalties in CUDA kernels, so consider using manual
                           cudaMemcpy operations from pinned host memory instead.
                        </p>
                        <p class="p"><strong class="ph b">PtoP transfer</strong> indicates the CUDA kernel accessed managed memory
                           that was residing on a different device, so the kernel execution paused
                           and transferred the data to this device.  Heavy traffic here will incur
                           performance penalties, so consider using manual cudaMemcpyPeer
                           operations to transfer from other devices' memory instead.  The row
                           showing these events is for the destination device -- the source device
                           is shown in the tooltip for each transfer event.
                        </p>
                        <p class="p"><strong class="ph b">DtoH transfer</strong> indicates the CPU accessed managed memory that was
                           residing on a CUDA device, so the CPU execution paused and transferred
                           the data to system memory. Heavy traffic here will incur performance
                           penalties in CPU code, so consider using manual cudaMemcpy operations
                           from pinned host memory instead.
                        </p>
                        <p class="p">
                           Some Unified Memory transfers are highlighted with red to indicate
                           		 potential performance issues:
                        </p>
                        <p class="p"><img class="image" src="graphics/uvm_migration_cause_hl.png" alt="Unified Memory transfer migration cause highlight"></img></p>
                        <div class="p">
                           Transfers with the following migration causes are highlighted:
                           
                           <ul class="ul">
                              <li class="li">
                                 <p class="p"><strong class="ph b">Coherence</strong></p>
                                 
                                 Unified Memory migration occurred to guarantee data coherence. SMs
                                 (streaming multiprocessors) stop until the migration completes.
                                 
                              </li>
                              <li class="li">
                                 <p class="p"><strong class="ph b">Eviction</strong></p>
                                 
                                 Unified Memory migrated to the CPU because it was evicted to make
                                 			 room for another block of memory on the GPU. This happens due to
                                 			 memory overcommitment which is available on Linux with
                                 Compute Capability ≥ 6.
                                 
                              </li>
                           </ul>
                        </div>
                     </div>
                     <div class="topic nested2" id="um-cpu-page-faults"><a name="um-cpu-page-faults" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#um-cpu-page-faults" name="um-cpu-page-faults" shape="rect">Unified Memory CPU Page Faults</a></h3>
                        <div class="body">
                           <p class="p">The Unified Memory CPU page faults feature in
                              			  <span class="ph">Nsight Systems</span> tracks the page faults that occur
                              			  when CPU code tries to access a memory page that resides on the
                              			  device.
                              			  
                           </p>
                           <blockquote class="lq">
                              <div class="p">
                                 <div class="tablenoborder">
                                    <table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
                                       <tbody class="tbody">
                                          <tr class="row">
                                             <td class="entry nvidiagreen" align="center" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                                <p class="p">&nbsp;&nbsp;<strong class="ph b">Note:</strong>&nbsp;&nbsp;
                                                </p>
                                             </td>
                                             <td class="entry notebox" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                                <p class="p">Collecting Unified Memory CPU page faults can cause
                                                   				   overhead of up to 70% in testing. Please use this functionality
                                                   				   only when needed.
                                                </p>
                                             </td>
                                          </tr>
                                       </tbody>
                                    </table>
                                 </div>
                              </div>
                           </blockquote>
                           <p class="p"><img class="image" src="graphics/uvm-cpu-page-faults.png" alt="Unified Memory CPU Page Faults on timeline"></img></p>
                        </div>
                     </div>
                     <div class="topic nested2" id="um-gpu-page-faults"><a name="um-gpu-page-faults" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#um-gpu-page-faults" name="um-gpu-page-faults" shape="rect">Unified Memory GPU Page Faults</a></h3>
                        <div class="body">
                           <p class="p">The Unified Memory GPU page faults feature in
                              			  <span class="ph">Nsight Systems</span> tracks the page faults that occur when
                              			  GPU code tries to access a memory page that resides on the host.
                              			  
                           </p>
                           <blockquote class="lq">
                              <div class="p">
                                 <div class="tablenoborder">
                                    <table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
                                       <tbody class="tbody">
                                          <tr class="row">
                                             <td class="entry nvidiagreen" align="center" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                                <p class="p">&nbsp;&nbsp;<strong class="ph b">Note:</strong>&nbsp;&nbsp;
                                                </p>
                                             </td>
                                             <td class="entry notebox" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                                <p class="p">Collecting Unified Memory GPU page faults can cause
                                                   				   overhead of up to 70% in testing. Please use this
                                                   				   functionality only when needed.
                                                </p>
                                             </td>
                                          </tr>
                                       </tbody>
                                    </table>
                                 </div>
                              </div>
                           </blockquote>
                           <p class="p"><img class="image" src="graphics/uvm-gpu-page-faults.png" alt="Unified Memory GPU Page Faults on timeline"></img></p>
                        </div>
                     </div>
                  </div>
                  <div class="topic nested1" id="cuda-graph"><a name="cuda-graph" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#cuda-graph" name="cuda-graph" shape="rect">15.3.&nbsp;CUDA Graph Trace</a></h3>
                     <div class="body">
                        <p class="p"><span class="ph">Nsight Systems</span> is capable of capturing information about CUDA
                           graphs in your application at either the graph or node granularity. This can
                           be set in the CLI using the <samp class="ph codeph">--cuda-graph-trace</samp> option, or
                           in the GUI by setting the appropriate drop down.
                        </p>
                        <blockquote class="lq">
                           <p class="p"><img class="image" src="graphics/cudagraph-selector.png" alt="Configure CUDA graph trace"></img></p>
                        </blockquote>
                        <p class="p">When CUDA graph trace is set to <samp class="ph codeph">graph</samp>, the users sees
                           each graph as one item on the timeline:
                        </p>
                        <blockquote class="lq">
                           <p class="p"><img class="image" src="graphics/cudagraph-graphlevel.png" alt="CUDA Graph trace at the graph level"></img></p>
                        </blockquote>
                        <p class="p">When CUDA graph trace is set to <samp class="ph codeph">node</samp>, the users sees
                           each graph as a set of nodes on the timeline:
                        </p>
                        <blockquote class="lq">
                           <p class="p"><img class="image" src="graphics/cudagraph-nodelevel.png" alt="CUDA Graph trace at the node level"></img></p>
                        </blockquote>
                        <p class="p">Tracing CUDA graphs at the graph level rather than the tracing the
                           underlying nodes results in significantly less overhead.
                           This option is only available with CUDA driver 515.43 or higher. 
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="cuda-default-cli"><a name="cuda-default-cli" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#cuda-default-cli" name="cuda-default-cli" shape="rect">15.4.&nbsp;CUDA Default Function List for CLI</a></h3>
                     <div class="body">
                        <p class="p"><strong class="ph b">CUDA Runtime API</strong></p><pre xml:space="preserve">cudaBindSurfaceToArray
cudaBindTexture
cudaBindTexture2D
cudaBindTextureToArray
cudaBindTextureToMipmappedArray
cudaConfigureCall
cudaCreateSurfaceObject
cudaCreateTextureObject
cudaD3D10MapResources
cudaD3D10RegisterResource
cudaD3D10UnmapResources
cudaD3D10UnregisterResource
cudaD3D9MapResources
cudaD3D9MapVertexBuffer
cudaD3D9RegisterResource
cudaD3D9RegisterVertexBuffer
cudaD3D9UnmapResources
cudaD3D9UnmapVertexBuffer
cudaD3D9UnregisterResource
cudaD3D9UnregisterVertexBuffer
cudaDestroySurfaceObject
cudaDestroyTextureObject
cudaDeviceReset
cudaDeviceSynchronize
cudaEGLStreamConsumerAcquireFrame
cudaEGLStreamConsumerConnect
cudaEGLStreamConsumerConnectWithFlags
cudaEGLStreamConsumerDisconnect
cudaEGLStreamConsumerReleaseFrame
cudaEGLStreamConsumerReleaseFrame
cudaEGLStreamProducerConnect
cudaEGLStreamProducerDisconnect
cudaEGLStreamProducerReturnFrame
cudaEventCreate
cudaEventCreateFromEGLSync
cudaEventCreateWithFlags
cudaEventDestroy
cudaEventQuery
cudaEventRecord
cudaEventRecord_ptsz
cudaEventSynchronize
cudaFree
cudaFreeArray
cudaFreeHost
cudaFreeMipmappedArray
cudaGLMapBufferObject
cudaGLMapBufferObjectAsync
cudaGLRegisterBufferObject
cudaGLUnmapBufferObject
cudaGLUnmapBufferObjectAsync
cudaGLUnregisterBufferObject
cudaGraphicsD3D10RegisterResource
cudaGraphicsD3D11RegisterResource
cudaGraphicsD3D9RegisterResource
cudaGraphicsEGLRegisterImage
cudaGraphicsGLRegisterBuffer
cudaGraphicsGLRegisterImage
cudaGraphicsMapResources
cudaGraphicsUnmapResources
cudaGraphicsUnregisterResource
cudaGraphicsVDPAURegisterOutputSurface
cudaGraphicsVDPAURegisterVideoSurface
cudaHostAlloc
cudaHostRegister
cudaHostUnregister
cudaLaunch
cudaLaunchCooperativeKernel
cudaLaunchCooperativeKernelMultiDevice
cudaLaunchCooperativeKernel_ptsz
cudaLaunchKernel
cudaLaunchKernel_ptsz
cudaLaunch_ptsz
cudaMalloc
cudaMalloc3D
cudaMalloc3DArray
cudaMallocArray
cudaMallocHost
cudaMallocManaged
cudaMallocMipmappedArray
cudaMallocPitch
cudaMemGetInfo
cudaMemPrefetchAsync
cudaMemPrefetchAsync_ptsz
cudaMemcpy
cudaMemcpy2D
cudaMemcpy2DArrayToArray
cudaMemcpy2DArrayToArray_ptds
cudaMemcpy2DAsync
cudaMemcpy2DAsync_ptsz
cudaMemcpy2DFromArray
cudaMemcpy2DFromArrayAsync
cudaMemcpy2DFromArrayAsync_ptsz
cudaMemcpy2DFromArray_ptds
cudaMemcpy2DToArray
cudaMemcpy2DToArrayAsync
cudaMemcpy2DToArrayAsync_ptsz
cudaMemcpy2DToArray_ptds
cudaMemcpy2D_ptds
cudaMemcpy3D
cudaMemcpy3DAsync
cudaMemcpy3DAsync_ptsz
cudaMemcpy3DPeer
cudaMemcpy3DPeerAsync
cudaMemcpy3DPeerAsync_ptsz
cudaMemcpy3DPeer_ptds
cudaMemcpy3D_ptds
cudaMemcpyArrayToArray
cudaMemcpyArrayToArray_ptds
cudaMemcpyAsync
cudaMemcpyAsync_ptsz
cudaMemcpyFromArray
cudaMemcpyFromArrayAsync
cudaMemcpyFromArrayAsync_ptsz
cudaMemcpyFromArray_ptds
cudaMemcpyFromSymbol
cudaMemcpyFromSymbolAsync
cudaMemcpyFromSymbolAsync_ptsz
cudaMemcpyFromSymbol_ptds
cudaMemcpyPeer
cudaMemcpyPeerAsync
cudaMemcpyToArray
cudaMemcpyToArrayAsync
cudaMemcpyToArrayAsync_ptsz
cudaMemcpyToArray_ptds
cudaMemcpyToSymbol
cudaMemcpyToSymbolAsync
cudaMemcpyToSymbolAsync_ptsz
cudaMemcpyToSymbol_ptds
cudaMemcpy_ptds
cudaMemset
cudaMemset2D
cudaMemset2DAsync
cudaMemset2DAsync_ptsz
cudaMemset2D_ptds
cudaMemset3D
cudaMemset3DAsync
cudaMemset3DAsync_ptsz
cudaMemset3D_ptds
cudaMemsetAsync
cudaMemsetAsync_ptsz
cudaMemset_ptds
cudaPeerRegister
cudaPeerUnregister
cudaStreamAddCallback
cudaStreamAddCallback_ptsz
cudaStreamAttachMemAsync
cudaStreamAttachMemAsync_ptsz
cudaStreamCreate
cudaStreamCreateWithFlags
cudaStreamCreateWithPriority
cudaStreamDestroy
cudaStreamQuery
cudaStreamQuery_ptsz
cudaStreamSynchronize
cudaStreamSynchronize_ptsz
cudaStreamWaitEvent
cudaStreamWaitEvent_ptsz
cudaThreadSynchronize
cudaUnbindTexture</pre><p class="p"><strong class="ph b">CUDA Primary API</strong></p><pre xml:space="preserve">cu64Array3DCreate
cu64ArrayCreate
cu64D3D9MapVertexBuffer
cu64GLMapBufferObject
cu64GLMapBufferObjectAsync
cu64MemAlloc
cu64MemAllocPitch
cu64MemFree
cu64MemGetInfo
cu64MemHostAlloc
cu64Memcpy2D
cu64Memcpy2DAsync
cu64Memcpy2DUnaligned
cu64Memcpy3D
cu64Memcpy3DAsync
cu64MemcpyAtoD
cu64MemcpyDtoA
cu64MemcpyDtoD
cu64MemcpyDtoDAsync
cu64MemcpyDtoH
cu64MemcpyDtoHAsync
cu64MemcpyHtoD
cu64MemcpyHtoDAsync
cu64MemsetD16
cu64MemsetD16Async
cu64MemsetD2D16
cu64MemsetD2D16Async
cu64MemsetD2D32
cu64MemsetD2D32Async
cu64MemsetD2D8
cu64MemsetD2D8Async
cu64MemsetD32
cu64MemsetD32Async
cu64MemsetD8
cu64MemsetD8Async
cuArray3DCreate
cuArray3DCreate_v2
cuArrayCreate
cuArrayCreate_v2
cuArrayDestroy
cuBinaryFree
cuCompilePtx
cuCtxCreate
cuCtxCreate_v2
cuCtxDestroy
cuCtxDestroy_v2
cuCtxSynchronize
cuD3D10CtxCreate
cuD3D10CtxCreateOnDevice
cuD3D10CtxCreate_v2
cuD3D10MapResources
cuD3D10RegisterResource
cuD3D10UnmapResources
cuD3D10UnregisterResource
cuD3D11CtxCreate
cuD3D11CtxCreateOnDevice
cuD3D11CtxCreate_v2
cuD3D9CtxCreate
cuD3D9CtxCreateOnDevice
cuD3D9CtxCreate_v2
cuD3D9MapResources
cuD3D9MapVertexBuffer
cuD3D9MapVertexBuffer_v2
cuD3D9RegisterResource
cuD3D9RegisterVertexBuffer
cuD3D9UnmapResources
cuD3D9UnmapVertexBuffer
cuD3D9UnregisterResource
cuD3D9UnregisterVertexBuffer
cuEGLStreamConsumerAcquireFrame
cuEGLStreamConsumerConnect
cuEGLStreamConsumerConnectWithFlags
cuEGLStreamConsumerDisconnect
cuEGLStreamConsumerReleaseFrame
cuEGLStreamProducerConnect
cuEGLStreamProducerDisconnect
cuEGLStreamProducerPresentFrame
cuEGLStreamProducerReturnFrame
cuEventCreate
cuEventCreateFromEGLSync
cuEventCreateFromNVNSync
cuEventDestroy
cuEventDestroy_v2
cuEventQuery
cuEventRecord
cuEventRecord_ptsz
cuEventSynchronize
cuGLCtxCreate
cuGLCtxCreate_v2
cuGLInit
cuGLMapBufferObject
cuGLMapBufferObjectAsync
cuGLMapBufferObjectAsync_v2
cuGLMapBufferObjectAsync_v2_ptsz
cuGLMapBufferObject_v2
cuGLMapBufferObject_v2_ptds
cuGLRegisterBufferObject
cuGLUnmapBufferObject
cuGLUnmapBufferObjectAsync
cuGLUnregisterBufferObject
cuGraphicsD3D10RegisterResource
cuGraphicsD3D11RegisterResource
cuGraphicsD3D9RegisterResource
cuGraphicsEGLRegisterImage
cuGraphicsGLRegisterBuffer
cuGraphicsGLRegisterImage
cuGraphicsMapResources
cuGraphicsMapResources_ptsz
cuGraphicsUnmapResources
cuGraphicsUnmapResources_ptsz
cuGraphicsUnregisterResource
cuGraphicsVDPAURegisterOutputSurface
cuGraphicsVDPAURegisterVideoSurface
cuInit
cuLaunch
cuLaunchCooperativeKernel
cuLaunchCooperativeKernelMultiDevice
cuLaunchCooperativeKernel_ptsz
cuLaunchGrid
cuLaunchGridAsync
cuLaunchKernel
cuLaunchKernel_ptsz
cuLinkComplete
cuLinkCreate
cuLinkCreate_v2
cuLinkDestroy
cuMemAlloc
cuMemAllocHost
cuMemAllocHost_v2
cuMemAllocManaged
cuMemAllocPitch
cuMemAllocPitch_v2
cuMemAlloc_v2
cuMemFree
cuMemFreeHost
cuMemFree_v2
cuMemGetInfo
cuMemGetInfo_v2
cuMemHostAlloc
cuMemHostAlloc_v2
cuMemHostRegister
cuMemHostRegister_v2
cuMemHostUnregister
cuMemPeerRegister
cuMemPeerUnregister
cuMemPrefetchAsync
cuMemPrefetchAsync_ptsz
cuMemcpy
cuMemcpy2D
cuMemcpy2DAsync
cuMemcpy2DAsync_v2
cuMemcpy2DAsync_v2_ptsz
cuMemcpy2DUnaligned
cuMemcpy2DUnaligned_v2
cuMemcpy2DUnaligned_v2_ptds
cuMemcpy2D_v2
cuMemcpy2D_v2_ptds
cuMemcpy3D
cuMemcpy3DAsync
cuMemcpy3DAsync_v2
cuMemcpy3DAsync_v2_ptsz
cuMemcpy3DPeer
cuMemcpy3DPeerAsync
cuMemcpy3DPeerAsync_ptsz
cuMemcpy3DPeer_ptds
cuMemcpy3D_v2
cuMemcpy3D_v2_ptds
cuMemcpyAsync
cuMemcpyAsync_ptsz
cuMemcpyAtoA
cuMemcpyAtoA_v2
cuMemcpyAtoA_v2_ptds
cuMemcpyAtoD
cuMemcpyAtoD_v2
cuMemcpyAtoD_v2_ptds
cuMemcpyAtoH
cuMemcpyAtoHAsync
cuMemcpyAtoHAsync_v2
cuMemcpyAtoHAsync_v2_ptsz
cuMemcpyAtoH_v2
cuMemcpyAtoH_v2_ptds
cuMemcpyDtoA
cuMemcpyDtoA_v2
cuMemcpyDtoA_v2_ptds
cuMemcpyDtoD
cuMemcpyDtoDAsync
cuMemcpyDtoDAsync_v2
cuMemcpyDtoDAsync_v2_ptsz
cuMemcpyDtoD_v2
cuMemcpyDtoD_v2_ptds
cuMemcpyDtoH
cuMemcpyDtoHAsync
cuMemcpyDtoHAsync_v2
cuMemcpyDtoHAsync_v2_ptsz
cuMemcpyDtoH_v2
cuMemcpyDtoH_v2_ptds
cuMemcpyHtoA
cuMemcpyHtoAAsync
cuMemcpyHtoAAsync_v2
cuMemcpyHtoAAsync_v2_ptsz
cuMemcpyHtoA_v2
cuMemcpyHtoA_v2_ptds
cuMemcpyHtoD
cuMemcpyHtoDAsync
cuMemcpyHtoDAsync_v2
cuMemcpyHtoDAsync_v2_ptsz
cuMemcpyHtoD_v2
cuMemcpyHtoD_v2_ptds
cuMemcpyPeer
cuMemcpyPeerAsync
cuMemcpyPeerAsync_ptsz
cuMemcpyPeer_ptds
cuMemcpy_ptds
cuMemcpy_v2
cuMemsetD16
cuMemsetD16Async
cuMemsetD16Async_ptsz
cuMemsetD16_v2
cuMemsetD16_v2_ptds
cuMemsetD2D16
cuMemsetD2D16Async
cuMemsetD2D16Async_ptsz
cuMemsetD2D16_v2
cuMemsetD2D16_v2_ptds
cuMemsetD2D32
cuMemsetD2D32Async
cuMemsetD2D32Async_ptsz
cuMemsetD2D32_v2
cuMemsetD2D32_v2_ptds
cuMemsetD2D8
cuMemsetD2D8Async
cuMemsetD2D8Async_ptsz
cuMemsetD2D8_v2
cuMemsetD2D8_v2_ptds
cuMemsetD32
cuMemsetD32Async
cuMemsetD32Async_ptsz
cuMemsetD32_v2
cuMemsetD32_v2_ptds
cuMemsetD8
cuMemsetD8Async
cuMemsetD8Async_ptsz
cuMemsetD8_v2
cuMemsetD8_v2_ptds
cuMipmappedArrayCreate
cuMipmappedArrayDestroy
cuModuleLoad
cuModuleLoadData
cuModuleLoadDataEx
cuModuleLoadFatBinary
cuModuleUnload
cuStreamAddCallback
cuStreamAddCallback_ptsz
cuStreamAttachMemAsync
cuStreamAttachMemAsync_ptsz
cuStreamBatchMemOp
cuStreamBatchMemOp_ptsz
cuStreamCreate
cuStreamCreateWithPriority
cuStreamDestroy
cuStreamDestroy_v2
cuStreamSynchronize
cuStreamSynchronize_ptsz
cuStreamWaitEvent
cuStreamWaitEvent_ptsz
cuStreamWaitValue32
cuStreamWaitValue32_ptsz
cuStreamWaitValue64
cuStreamWaitValue64_ptsz
cuStreamWriteValue32
cuStreamWriteValue32_ptsz
cuStreamWriteValue64
cuStreamWriteValue64_ptsz
cuSurfObjectCreate
cuSurfObjectDestroy
cuSurfRefCreate
cuSurfRefDestroy
cuTexObjectCreate
cuTexObjectDestroy
cuTexRefCreate
cuTexRefDestroy
cuVDPAUCtxCreate
cuVDPAUCtxCreate_v2</pre></div>
                  </div>
                  <div class="topic nested1" id="cudnn-function-cli"><a name="cudnn-function-cli" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#cudnn-function-cli" name="cudnn-function-cli" shape="rect">15.5.&nbsp;cuDNN Function List for X86 CLI</a></h3>
                     <div class="body">
                        <p class="p"><strong class="ph b">cuDNN API functions</strong></p><pre xml:space="preserve">cudnnActivationBackward
cudnnActivationBackward_v3
cudnnActivationBackward_v4
cudnnActivationForward
cudnnActivationForward_v3
cudnnActivationForward_v4
cudnnAddTensor
cudnnBatchNormalizationBackward
cudnnBatchNormalizationBackwardEx
cudnnBatchNormalizationForwardInference
cudnnBatchNormalizationForwardTraining
cudnnBatchNormalizationForwardTrainingEx
cudnnCTCLoss
cudnnConvolutionBackwardBias
cudnnConvolutionBackwardData
cudnnConvolutionBackwardFilter
cudnnConvolutionBiasActivationForward
cudnnConvolutionForward
cudnnCreate
cudnnCreateAlgorithmPerformance
cudnnDestroy
cudnnDestroyAlgorithmPerformance
cudnnDestroyPersistentRNNPlan
cudnnDivisiveNormalizationBackward
cudnnDivisiveNormalizationForward
cudnnDropoutBackward
cudnnDropoutForward
cudnnDropoutGetReserveSpaceSize
cudnnDropoutGetStatesSize
cudnnFindConvolutionBackwardDataAlgorithm
cudnnFindConvolutionBackwardDataAlgorithmEx
cudnnFindConvolutionBackwardFilterAlgorithm
cudnnFindConvolutionBackwardFilterAlgorithmEx
cudnnFindConvolutionForwardAlgorithm
cudnnFindConvolutionForwardAlgorithmEx
cudnnFindRNNBackwardDataAlgorithmEx
cudnnFindRNNBackwardWeightsAlgorithmEx
cudnnFindRNNForwardInferenceAlgorithmEx
cudnnFindRNNForwardTrainingAlgorithmEx
cudnnFusedOpsExecute
cudnnIm2Col
cudnnLRNCrossChannelBackward
cudnnLRNCrossChannelForward
cudnnMakeFusedOpsPlan
cudnnMultiHeadAttnBackwardData
cudnnMultiHeadAttnBackwardWeights
cudnnMultiHeadAttnForward
cudnnOpTensor
cudnnPoolingBackward
cudnnPoolingForward
cudnnRNNBackwardData
cudnnRNNBackwardDataEx
cudnnRNNBackwardWeights
cudnnRNNBackwardWeightsEx
cudnnRNNForwardInference
cudnnRNNForwardInferenceEx
cudnnRNNForwardTraining
cudnnRNNForwardTrainingEx
cudnnReduceTensor
cudnnReorderFilterAndBias
cudnnRestoreAlgorithm
cudnnRestoreDropoutDescriptor
cudnnSaveAlgorithm
cudnnScaleTensor
cudnnSoftmaxBackward
cudnnSoftmaxForward
cudnnSpatialTfGridGeneratorBackward
cudnnSpatialTfGridGeneratorForward
cudnnSpatialTfSamplerBackward
cudnnSpatialTfSamplerForward
cudnnTransformFilter
cudnnTransformTensor
cudnnTransformTensorEx</pre><p class="p"></p>
                        <p class="p hr"></p>
                        <p class="p"></p>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="openacc-trace"><a name="openacc-trace" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#openacc-trace" name="openacc-trace" shape="rect">16.&nbsp;OpenACC Trace</a></h2>
                  <div class="body">
                     <p class="p"><span class="ph">Nsight Systems</span> for Linux x86_64 and Power targets is capable
                        of capturing information about OpenACC execution in the profiled process.
                     </p>
                     <p class="p">
                        OpenACC versions 2.0, 2.5, and 2.6 are supported when using PGI runtime
                        version 15.7 or later. In order to differentiate constructs (see tooltip
                        below), a PGI runtime of 16.0 or later is required. Note that
                        <span class="ph">Nsight Systems</span> does not support the GCC implementation of
                        OpenACC at this time.
                     </p>
                     <p class="p">Under the CPU rows in the timeline tree, each thread that uses OpenACC
                        will show OpenACC trace information. You can click on a OpenACC API call to
                        see correlation with the underlying CUDA API calls (highlighted in teal):
                     </p>
                     <p class="p"><img class="image" src="graphics/openacc-CUDA-correlation.png" alt="OpenACC rows"></img></p>
                     <p class="p">If the OpenACC API results in GPU work, that will also be highlighted:</p>
                     <p class="p"><img class="image" src="graphics/openacc-gpu-correlation.png" alt="OpenACC rows"></img></p>
                     <p class="p">Hovering over a particular OpenACC construct will bring up a tooltip with 
                        details about that construct:
                     </p>
                     <blockquote class="lq">
                        <p class="p"><img class="image" src="graphics/openacc-tooltip.png" alt="OpenACC construct tooltip"></img></p>
                     </blockquote>
                     <p class="p">
                        To capture OpenACC information from the <span class="ph">Nsight Systems</span> GUI,
                        select the <strong class="ph b">Collect OpenACC trace</strong> checkbox under <strong class="ph b">Collect CUDA
                           trace</strong> configurations. Note that turning on OpenACC tracing will also
                        turn on CUDA tracing.
                     </p>
                     <blockquote class="lq">
                        <p class="p"><img class="image" src="graphics/configure-cuda-trace-x86.png" alt="Configure CUDA trace"></img></p>
                     </blockquote>
                     <p class="p">
                        Please note that if your application crashes before all collected OpenACC
                        trace data has been copied out, some or all data might be lost and not
                        present in the report.
                     </p>
                     <p class="p"></p>
                     <p class="p hr"></p>
                     <p class="p"></p>
                  </div>
               </div>
               <div class="topic nested0" id="opengl-trace"><a name="opengl-trace" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#opengl-trace" name="opengl-trace" shape="rect">17.&nbsp;OpenGL Trace</a></h2>
                  <div class="body">
                     <p class="p">OpenGL and OpenGL ES APIs can be traced to assist in the analysis of CPU and
                        GPU interactions.
                     </p>
                     <p class="p">A few usage examples are:</p>
                     <ol class="ol">
                        <li class="li">
                           <p class="p">Visualize how long <samp class="ph codeph">eglSwapBuffers</samp> (or similar) is taking.
                           </p>
                        </li>
                        <li class="li">
                           <p class="p">API trace can easily show correlations between thread state and graphics
                              driver's behavior, uncovering where the CPU may be waiting on the GPU.
                           </p>
                        </li>
                        <li class="li">
                           <p class="p">Spot bubbles of opportunity on the GPU, where more GPU workload could be
                              created.
                           </p>
                        </li>
                        <li class="li">
                           <p class="p">Use <samp class="ph codeph">KHR_debug</samp> extension to trace GL events on both the CPU and GPU.
                           </p>
                        </li>
                     </ol>
                     <p class="p">OpenGL trace feature in <span class="ph">Nsight Systems</span> consists of two
                        different activities which will be shown in the CPU rows for those threads
                     </p>
                     <ul class="ul">
                        <li class="li">
                           <p class="p"><strong class="ph b">CPU trace</strong>: interception of API calls that an application does to APIs
                              (such as OpenGL, OpenGL ES, EGL, GLX, WGL, etc.).
                           </p>
                        </li>
                        <li class="li">
                           <p class="p"><strong class="ph b">GPU trace</strong> (or <strong class="ph b">workload trace</strong>): trace of GPU workload (activity)
                              triggered by use of OpenGL or OpenGL ES. Since draw calls are executed
                              back-to-back, the GPU workload trace ranges include many OpenGL draw calls
                              and operations in order to optimize performance overhead, rather than
                              tracing each individual operation.
                           </p>
                        </li>
                     </ul>
                     <p class="p">
                        To collect GPU trace, the <samp class="ph codeph">glQueryCounter()</samp> function is used
                        to measure how much time batches of GPU workload take to complete.
                     </p>
                     <blockquote class="lq">
                        <p class="p"><img class="image" src="graphics/configure-opengl-trace.png" alt="Configure OpenGL trace"></img></p>
                     </blockquote>
                     <blockquote class="lq">
                        <p class="p"><img class="image" src="graphics/configure-opengl-functions.png" alt="Configure OpenGL functions"></img></p>
                     </blockquote>
                     <p class="p">
                        Ranges defined by the <samp class="ph codeph">KHR_debug</samp> calls are represented
                        similarly to OpenGL API and OpenGL GPU workload trace. GPU ranges in this
                        case represent <em class="ph i">incremental draw cost</em>.  They cannot fully account for
                        GPUs that can execute multiple draw calls in parallel.  In this case,
                        <span class="ph">Nsight Systems</span> will not show overlapping GPU ranges.
                     </p>
                  </div>
                  <div class="topic nested1" id="opengl-cli"><a name="opengl-cli" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#opengl-cli" name="opengl-cli" shape="rect">17.1.&nbsp;OpenGL Trace Using Command Line</a></h3>
                     <div class="body">
                        <p class="p">For general information on using the target CLI, see
                           <a class="xref" href="topics/cli-profiling" shape="rect">CLI Profiling on Linux</a>. For the
                           CLI, the functions that are traced are set to the following list:
                        </p><pre xml:space="preserve">glWaitSync
glReadPixels
glReadnPixelsKHR
glReadnPixelsEXT
glReadnPixelsARB
glReadnPixels
glFlush
glFinishFenceNV
glFinish
glClientWaitSync
glClearTexSubImage
glClearTexImage
glClearStencil
glClearNamedFramebufferuiv
glClearNamedFramebufferiv
glClearNamedFramebufferfv
glClearNamedFramebufferfi
glClearNamedBufferSubDataEXT
glClearNamedBufferSubData
glClearNamedBufferDataEXT
glClearNamedBufferData
glClearIndex
glClearDepthx
glClearDepthf
glClearDepthdNV
glClearDepth
glClearColorx
glClearColorIuiEXT
glClearColorIiEXT
glClearColor
glClearBufferuiv
glClearBufferSubData
glClearBufferiv
glClearBufferfv
glClearBufferfi
glClearBufferData
glClearAccum
glClear
glDispatchComputeIndirect
glDispatchComputeGroupSizeARB
glDispatchCompute
glComputeStreamNV
glNamedFramebufferDrawBuffers
glNamedFramebufferDrawBuffer
glMultiDrawElementsIndirectEXT
glMultiDrawElementsIndirectCountARB
glMultiDrawElementsIndirectBindlessNV
glMultiDrawElementsIndirectBindlessCountNV
glMultiDrawElementsIndirectAMD
glMultiDrawElementsIndirect
glMultiDrawElementsEXT
glMultiDrawElementsBaseVertex
glMultiDrawElements
glMultiDrawArraysIndirectEXT
glMultiDrawArraysIndirectCountARB
glMultiDrawArraysIndirectBindlessNV
glMultiDrawArraysIndirectBindlessCountNV
glMultiDrawArraysIndirectAMD
glMultiDrawArraysIndirect
glMultiDrawArraysEXT
glMultiDrawArrays
glListDrawCommandsStatesClientNV
glFramebufferDrawBuffersEXT
glFramebufferDrawBufferEXT
glDrawTransformFeedbackStreamInstanced
glDrawTransformFeedbackStream
glDrawTransformFeedbackNV
glDrawTransformFeedbackInstancedEXT
glDrawTransformFeedbackInstanced
glDrawTransformFeedbackEXT
glDrawTransformFeedback
glDrawTexxvOES
glDrawTexxOES
glDrawTextureNV
glDrawTexsvOES
glDrawTexsOES
glDrawTexivOES
glDrawTexiOES
glDrawTexfvOES
glDrawTexfOES
glDrawRangeElementsEXT
glDrawRangeElementsBaseVertexOES
glDrawRangeElementsBaseVertexEXT
glDrawRangeElementsBaseVertex
glDrawRangeElements
glDrawPixels
glDrawElementsInstancedNV
glDrawElementsInstancedEXT
glDrawElementsInstancedBaseVertexOES
glDrawElementsInstancedBaseVertexEXT
glDrawElementsInstancedBaseVertexBaseInstanceEXT
glDrawElementsInstancedBaseVertexBaseInstance
glDrawElementsInstancedBaseVertex
glDrawElementsInstancedBaseInstanceEXT
glDrawElementsInstancedBaseInstance
glDrawElementsInstancedARB
glDrawElementsInstanced
glDrawElementsIndirect
glDrawElementsBaseVertexOES
glDrawElementsBaseVertexEXT
glDrawElementsBaseVertex
glDrawElements
glDrawCommandsStatesNV
glDrawCommandsStatesAddressNV
glDrawCommandsNV
glDrawCommandsAddressNV
glDrawBuffersNV
glDrawBuffersATI
glDrawBuffersARB
glDrawBuffers
glDrawBuffer
glDrawArraysInstancedNV
glDrawArraysInstancedEXT
glDrawArraysInstancedBaseInstanceEXT
glDrawArraysInstancedBaseInstance
glDrawArraysInstancedARB
glDrawArraysInstanced
glDrawArraysIndirect
glDrawArraysEXT
glDrawArrays
eglSwapBuffersWithDamageKHR
eglSwapBuffers
glXSwapBuffers
glXQueryDrawable
glXGetCurrentReadDrawable
glXGetCurrentDrawable
glGetQueryObjectuivEXT
glGetQueryObjectuivARB
glGetQueryObjectuiv
glGetQueryObjectivARB
glGetQueryObjectiv</pre><p class="p"></p>
                        <p class="p hr"></p>
                        <p class="p"></p>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="custom-etw-trace"><a name="custom-etw-trace" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#custom-etw-trace" name="custom-etw-trace" shape="rect">18.&nbsp;Custom ETW Trace</a></h2>
                  <div class="body">
                     <p class="p">
                        Use the custom ETW trace feature to enable and collect any manifest-based
                        ETW log. The collected events are displayed on the timeline on dedicated
                        rows for each event type.
                     </p>
                     <p class="p">
                        Custom ETW is available on Windows target machines.
                     </p>
                     <blockquote class="lq">
                        <p class="p"><img class="image" src="graphics/custom-etw-add-provider.png" alt="Adding details of an ETW provider"></img></p>
                     </blockquote>
                     <blockquote class="lq">
                        <p class="p"><img class="image" src="graphics/custom-etw-trace-settings.png" alt="Adding an ETW provider to the trace settings"></img></p>
                     </blockquote>
                     <p class="p"><img class="image" src="graphics/custom-etw-events-timeline.png" alt="Display of custom ETW trace events on the timeline"></img></p>
                     <p class="p">To retain the .etl trace files captured, so that they can be viewed in
                        other tools (e.g. GPUView), change the "Save ETW log files in project
                        folder" option under "Profile Behavior" in <span class="ph">Nsight Systems</span>'s
                        global Options dialog. The .etl files will appear in the same folder as the
                        .nsys-rep file, accessible by right-clicking the report in the Project Explorer
                        and choosing "Show in Folder...". Data collected from each ETW provider
                        will appear in its own .etl file, and an additional .etl file named
                        "Report XX-Merged-*.etl", containing the events from all captured sources,
                        will be created as well.
                        
                     </p>
                     <p class="p"></p>
                     <p class="p hr"></p>
                     <p class="p"></p>
                  </div>
               </div>
               <div class="topic nested0" id="gpu-metric-sampling"><a name="gpu-metric-sampling" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#gpu-metric-sampling" name="gpu-metric-sampling" shape="rect">GPU Metrics</a></h2>
                  <div class="body"></div>
                  <div class="topic nested1" id="gms-introduction"><a name="gms-introduction" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#gms-introduction" name="gms-introduction" shape="rect">Overview</a></h3>
                     <div class="body">
                        <p class="p">GPU Metrics feature is intended to identify performance limiters in
                           applications using GPU for computations and graphics. It uses periodic
                           sampling to gather performance metrics and detailed timing statistics
                           associated with different GPU hardware units taking advantage of
                           specialized hardware to capture this data in a single pass with
                           minimal overhead.
                        </p>
                        <p class="p"><strong class="ph b">Note:</strong> GPU Metrics will give you precise device level information,
                           but it does not know which process or context is involved. GPU context switch trace provides
                           less precise information, but will give you process and context information.
                        </p>
                        <p class="p"><img class="image" src="graphics/gpumetrics_screenshot.png" alt="Example GPU Metrics screenshot"></img></p>
                        <p class="p">These metrics provide an overview of GPU efficiency over time within
                           compute, graphics, and input/output (IO) activities such as:
                        </p>
                        <ul class="ul">
                           <li class="li"><strong class="ph b">IO throughputs:</strong>  PCIe, NVLink, and GPU memory bandwidth
                           </li>
                           <li class="li"><strong class="ph b">SM utilization:</strong>  SMs activity, tensor core activity, instructions
                              issued, warp occupancy, and unassigned warp slots
                           </li>
                        </ul>
                        <p class="p">It is designed to help users answer the common questions:</p>
                        <ul class="ul">
                           <li class="li">Is my GPU idle?</li>
                           <li class="li">Is my GPU full? Enough kernel grids size and streams? Are my SMs and
                              warp slots full?
                           </li>
                           <li class="li">Am I using TensorCores?</li>
                           <li class="li">Is my instruction rate high?</li>
                           <li class="li">Am I possibly blocked on IO, or number of warps, etc</li>
                        </ul>
                        <p class="p"><span class="ph">Nsight Systems</span> GPU Metrics is only available for Linux
                           targets on x86-64 and aarch64, and for Windows targets. It requires
                           NVIDIA Turing architecture or newer.
                        </p>
                        <div class="p">Minimum required driver versions:
                           
                           <ul class="ul">
                              <li class="li">NVIDIA Turing architecture TU10x, TU11x - r440</li>
                              <li class="li">NVIDIA Ampere architecture GA100 - r450</li>
                              <li class="li">NVIDIA Ampere architecture GA100 MIG - r470 TRD1</li>
                              <li class="li">NVIDIA Ampere architecture GA10x - r455</li>
                           </ul>
                        </div>
                        <p class="p"><strong class="ph b">Note:</strong> Elevated permissions are required. On Linux use sudo to
                           elevate privileges. On Windows the user must run from an admin command
                           prompt or accept the UAC escalation dialog. See
                           <a class="xref" href="https://developer.nvidia.com/ERR_NVGPUCTRPERM" target="_blank" shape="rect">Permissions Issues and Performance Counters</a>
                           for more information.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="gms-from-CLI"><a name="gms-from-CLI" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#gms-from-CLI" name="gms-from-CLI" shape="rect">Launching GPU Metric from the CLI</a></h3>
                     <div class="body">
                        <div class="p">GPU Metrics feature is controlled with 3 CLI switches:
                           
                           <ul class="ul">
                              <li class="li">--gpu-metrics-device=[all, none, &lt;index&gt;] selects GPUs to sample (default is none)</li>
                              <li class="li">--gpu-metrics-set=[&lt;index&gt;, &lt;alias&gt;] selects metric set to use (default is the 1st suitable from the list)</li>
                              <li class="li">--gpu-metrics-frequency=[10..200000] selects sampling frequency in Hz (default is 10000)</li>
                           </ul>
                        </div>
                        <div class="p">To profile with default options and sample GPU Metrics on GPU 0:
                           <pre xml:space="preserve"># Must have elevated permissions (see https://developer.nvidia.com/ERR_NVGPUCTRPERM) or be root (Linux) or Administrator (Windows)
$ <span class="ph">nsys</span> profile --gpu-metrics-device=0 ./my-app</pre></div>
                        <div class="p">To list available GPUs, use:
                           <pre xml:space="preserve">$ <span class="ph">nsys</span> profile --gpu-metrics-device=help
Possible --gpu-metrics-device values are:
    0: Quadro GV100 PCI[0000:17:00.0]
    1: GeForce RTX 2070 SUPER PCI[0000:65:00.0]
    all: Select all supported GPUs
    none: Disable GPU Metrics [Default]</pre></div>
                        <div class="p">By default, the first <strong class="ph b">metric set</strong> which supports all selected
                           GPUs is used. But you can manually select another metric set from the
                           list. To see available metric sets, use:
                           <pre xml:space="preserve">$ <span class="ph">nsys</span> profile --gpu-metrics-set=help
Possible --gpu-metrics-set values are:
    [0] [tu10x]        General Metrics for NVIDIA TU10x (any frequency)
    [1] [tu11x]        General Metrics for NVIDIA TU11x (any frequency)
    [2] [ga100]        General Metrics for NVIDIA GA100 (any frequency)
    [3] [ga10x]        General Metrics for NVIDIA GA10x (any frequency)
    [4] [tu10x-gfxt]   Graphics Throughput Metrics for NVIDIA TU10x (frequency &gt;= 10kHz)
    [5] [ga10x-gfxt]   Graphics Throughput Metrics for NVIDIA GA10x (frequency &gt;= 10kHz)
    [6] [ga10x-gfxact] Graphics Async Compute Triage Metrics for NVIDIA GA10x (frequency &gt;= 10kHz)</pre></div>
                        <div class="p">By default, <strong class="ph b">sampling frequency</strong> is set to 10 kHz. But you can manually set it from 10 Hz to 200 kHz using
                           <pre xml:space="preserve">--gpu-metrics-frequency=&lt;value&gt;</pre></div>
                     </div>
                  </div>
                  <div class="topic nested1" id="gms-from-GUI"><a name="gms-from-GUI" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#gms-from-GUI" name="gms-from-GUI" shape="rect">Launching GPU Metrics from the GUI</a></h3>
                     <div class="body">
                        <p class="p">For commands to launch GPU Metrics from the CLI with examples, see the
                           <a class="xref" href="topics/cli-profiling" shape="rect">CLI documentation</a>.
                        </p>
                        <p class="p"> When launching analysis in <span class="ph">Nsight Systems</span>, select
                           <strong class="ph b">Collect GPU Metrics</strong>.
                        </p>
                        <p class="p"><img class="image" src="graphics/gpumetrics-GUI.png" alt="GPU Metrics: GUI"></img></p>
                        <p class="p"> Select the <strong class="ph b">GPUs</strong> dropdown to pick which GPUs you wish to
                           sample.
                        </p>
                        <p class="p"> Select the <strong class="ph b">Metric set:</strong> dropdown to choose which available
                           metric set you would like to sample.
                        </p>
                        <p class="p"><img class="image" src="graphics/gpumetrics-GUI-2.png" alt="GPU Metrics: Metric sets"></img></p>
                        <p class="p">Note that metric sets for GPUs that are not being sampled will be
                           greyed out.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="gms-sampling-rate"><a name="gms-sampling-rate" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#gms-sampling-rate" name="gms-sampling-rate" shape="rect">Sampling frequency</a></h3>
                     <div class="body">
                        <p class="p">Sampling frequency can be selected from the range of 10 Hz - 200 kHz.
                           The default value is 10 kHz.
                        </p>
                        <p class="p">The maximum sampling frequency without buffer overflow events depends
                           on GPU (SM count), GPU load intensity, and overall system load. The
                           bigger the chip and the higher the load, the lower the maximum frequency.
                           If you need higher frequency, you can increase it until you get
                           "Buffer overflow" message in the Diagnostics Summary report page.
                        </p>
                        <p class="p">Each metric set has a recommended sampling frequency range in its
                           description. These ranges are approximate. If you observe
                           <samp class="ph codeph">Inonsistent Data</samp> or <samp class="ph codeph">Missing Data</samp> ranges
                           on timeline, please try closer to the recommended frequency.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="gms-available-metrics"><a name="gms-available-metrics" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#gms-available-metrics" name="gms-available-metrics" shape="rect">Available metrics</a></h3>
                     <div class="body">
                        <ul class="ul">
                           <li class="li"><strong class="ph b">GPC Clock Frequency</strong> -
                              <samp class="ph codeph">gpc__cycles_elapsed.avg.per_second</samp><p class="p">The average GPC clock frequency in hertz. In public documentation
                                 the GPC clock may be called the "Application" clock, "Graphic"
                                 clock, "Base" clock, or "Boost" clock.
                              </p>
                              <p class="p"><strong class="ph b">Note</strong>: The collection mechanism for GPC can result in a
                                 small fluctuation between samples.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">SYS Clock Frequency</strong> -
                              <samp class="ph codeph">sys__cycles_elapsed.avg.per_second</samp><p class="p">The average SYS clock frequency in hertz. The GPU front end
                                 (command processor), copy engines, and the performance monitor run
                                 at the SYS clock. On Turing and NVIDIA GA100 GPUs the sampling
                                 frequency is based upon a period of SYS clocks (not time)
                                 so samples per second will vary with SYS clock. On NVIDIA GA10x GPUs
                                 the sampling frequency is based upon a fixed frequency clock.
                                 The maximum frequency scales linearly with the SYS clock.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">GR Active</strong> -
                              <samp class="ph codeph">gr__cycles_active.sum.pct_of_peak_sustained_elapsed</samp><p class="p">The percentage of cycles the graphics/compute engine is active. The
                                 graphics/compute engine is active if there is any work in the graphics
                                 pipe or if the compute pipe is processing work.
                              </p>
                              <p class="p">GA100 MIG - MIG is not yet supported. This counter will report the
                                 activity of the primary GR engine.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">Sync Compute In Flight</strong> -
                              <samp class="ph codeph">gr__dispatch_cycles_active_queue_sync.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The percentage of cycles with synchronous compute in flight.</p>
                              <p class="p">CUDA: CUDA will only report synchronous queue in the case of MPS
                                 configured with 64 sub-context. Synchronous refers to work submitted
                                 in VEID=0.
                              </p>
                              <p class="p">Graphics: This will be true if any compute work submitted from the
                                 direct queue is in flight.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">Async Compute in Flight</strong> -
                              <samp class="ph codeph">gr__dispatch_cycles_active_queue_async.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The percentage of cycles with asynchronous compute in flight.</p>
                              <p class="p">CUDA: CUDA will only report all compute work as asynchronous. The
                                 one exception is if MPS is configured and all 64 sub-context are in
                                 use. 1 sub-context (VEID=0) will report as synchronous.
                              </p>
                              <p class="p">Graphics: This will be true if any compute work submitted from a
                                 compute queue is in flight.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">Draw Started</strong> -
                              <samp class="ph codeph">fe__draw_count.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The ratio of draw calls issued to the graphics pipe to the maximum
                                 sustained rate of the graphics pipe.
                              </p>
                              <p class="p"><strong class="ph b">Note:</strong>The percentage will always be very low as the front end
                                 can issue draw calls significantly faster than the pipe can execute
                                 the draw call. The rendering of this row will be changed to help
                                 indicate when draw calls are being issued.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">Dispatch Started</strong> -
                              <samp class="ph codeph">gr__dispatch_count.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The ratio of compute grid launches (dispatches) to the compute pipe
                                 to the maximum sustained rate of the compute pipe.
                              </p>
                              <p class="p"><strong class="ph b">Note:</strong> The percentage will always be very low as the front
                                 end can issue grid launches significantly faster than the pipe can
                                 execute the draw call. The rendering of this row will be changed to
                                 help indicate when grid launches are being issued.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">Vertex/Tess/Geometry Warps in Flight</strong> -
                              <samp class="ph codeph">tpc__warps_active_shader_vtg_realtime.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The ratio of active vertex, geometry, tessellation, and meshlet
                                 shader warps resident on the SMs to the maximum number of warps per SM
                                 as a percentage.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">Pixel Warps in Flight</strong> -
                              <samp class="ph codeph">tpc__warps_active_shader_ps_realtime.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The ratio of active pixel/fragment shader warps resident on the SMs
                                 to the maximum number of warps per SM as a percentage.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">Compute Warps in Flight</strong> -
                              <samp class="ph codeph">tpc__warps_active_shader_cs_realtime.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The ratio of active compute shader warps resident on the SMs to the
                                 maximum number of warps per SM as a percentage.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">Active SM Unused Warp Slots</strong> -
                              <samp class="ph codeph">tpc__warps_inactive_sm_active_realtime.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The ratio of inactive warp slots on the SMs to the maximum number
                                 of warps per SM as a percentage. This is an indication of how many
                                 more warps may fit on the SMs if occupancy is not limited by a resource
                                 such as max warps of a shader type, shared memory, registers per
                                 thread, or thread blocks per SM.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">Idle SM Unused Warp Slots</strong> -
                              <samp class="ph codeph">tpc__warps_inactive_sm_idle_realtime.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The ratio of inactive warps slots due to idle SMs to the the maximum
                                 number of warps per SM as a percentage.
                              </p>
                              <div class="p">This is an indicator that the current workload on the SM is not
                                 sufficient to put work on all SMs. This can be due to:
                                 
                                 <ul class="ul">
                                    <li class="li">CPU starving the GPU</li>
                                    <li class="li">current work is too small to saturate the GPU</li>
                                    <li class="li">current work is trailing off but blocking next work</li>
                                 </ul>
                              </div>
                           </li>
                           <li class="li"><strong class="ph b">SM Active</strong> -
                              <samp class="ph codeph">sm__cycles_active.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The ratio of cycles SMs had at least 1 warp in flight (allocated on
                                 SM) to the number of cycles as a percentage. A value of 0 indicates
                                 all SMs were idle (no warps in flight). A value of 50% can indicate
                                 some gradient between all SMs active 50% of the sample period or 50%
                                 of SMs active 100% of the sample period.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">SM Issue</strong> -
                              <samp class="ph codeph">sm__inst_executed_realtime.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The ratio of cycles that SM sub-partitions (warp schedulers) issued
                                 an instruction to the number of cycles in the sample period as a
                                 percentage.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">Tensor Active</strong> -
                              <samp class="ph codeph">sm__pipe_tensor_cycles_active_realtime.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The ratio of cycles the SM tensor pipes were active issuing tensor
                                 instructions to the number of cycles in the sample period as a percentage.
                              </p>
                              <p class="p">TU102/4/6: This metric is not available on TU10x for periodic
                                 sampling. Please see Tensor Active/FP16 Active.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">Tensor Active / FP16 Active</strong> -
                              <samp class="ph codeph">sm__pipe_shared_cycles_active_realtime.avg.pct_of_peak_sustained_elapsed</samp><p class="p">TU102/4/6 only</p>
                              <p class="p">The ratio of cycles the SM tensor pipes or FP16x2 pipes were active
                                 issuing tensor instructions to the number of cycles in the sample
                                 period as a percentage.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">VRAM Bandwidth</strong> -
                              <samp class="ph codeph">dram__throughput.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The ratio of cycles the GPU device memory controllers were actively
                                 performing read or write operations to the number of cycles in the
                                 sample period as a percentage.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">NVLink bytes received</strong> -
                              <samp class="ph codeph">nvlrx__bytes.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The ratio of bytes received on the NVLink interface to the maximum
                                 number of bytes receivable in the sample period as a percentage. This
                                 value includes protocol overhead.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">NVLink bytes transmitted</strong> -
                              <samp class="ph codeph">nvltx__bytes.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The ratio of bytes transmitted on the NVLink interface to the
                                 maximum number of bytes transmittable in the sample period as a
                                 percentage. This value includes protocol overhead.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">PCIe Read Throughput</strong> -
                              <samp class="ph codeph">pcie__read_bytes.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The ratio of bytes received on the PCIe interface to the maximum
                                 number of bytes receivable in the sample period as a percentage. The
                                 theoretical value is calculated based upon the PCIe generation and
                                 number of lanes. This value includes protocol overhead.
                              </p>
                           </li>
                           <li class="li"><strong class="ph b">PCIe Write Throughput</strong> -
                              <samp class="ph codeph">pcie__write_bytes.avg.pct_of_peak_sustained_elapsed</samp><p class="p">The ratio of bytes received on the PCIe interface to the maximum
                                 number of bytes receivable in the sample period as a percentage. The
                                 theoretical value is calculated based upon the PCIe generation and
                                 number of lanes. This value includes protocol overhead.
                              </p>
                           </li>
                        </ul>
                     </div>
                  </div>
                  <div class="topic nested1" id="gms-export-query"><a name="gms-export-query" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#gms-export-query" name="gms-export-query" shape="rect">Exporting and Querying Data</a></h3>
                     <div class="body">
                        <p class="p">It is possible to access metric values for automated processing
                           using the <span class="ph">Nsight Systems</span> CLI export capabilities.
                        </p>
                        <p class="p">An example that extracts values of "SM Active":</p><pre xml:space="preserve">
      $ <span class="ph">nsys</span> export -t sqlite report.nsys-rep
      $ sqlite3 report.sqlite "SELECT rawTimestamp, CAST(JSON_EXTRACT(data, '$.\"SM Active\"') as INTEGER) as value FROM GENERIC_EVENTS WHERE value != 0 LIMIT 10"

      309277039|80
      309301295|99
      309325583|99
      309349776|99
      309373872|60
      309397872|19
      309421840|100
      309446000|100
      309470096|100
      309494161|99
      </pre><p class="p">An overview of data stored in each event (JSON):</p><pre xml:space="preserve">
      $ sqlite3 report.sqlite "SELECT data FROM GENERIC_EVENTS LIMIT 1"
      {
      "Unallocated Warps in Active SM": "0",
      "Compute Warps In Flight": "52",
      "Pixel Warps In Flight": "0",
      "Vertex\/Tess\/Geometry Warps In Flight": "0",
      "Total SM Occupancy": "52",
      "GR Active (GE\/CE)": "100",
      "Sync Compute In Flight": "0",
      "Async Compute In Flight": "98",
      "NVLink bytes received": "0",
      "NVLink bytes transmitted": "0",
      "PCIe Rx Throughput": "0",
      "PCIe Tx Throughput": "1",
      "DRAM Read Throughput": "0",
      "DRAM Write Throughput": "0",
      "Tensor Active \/ FP16 Active": "0",
      "SM Issue": "10",
      "SM Active": "52"
      }
      </pre><p class="p">Values are integer percentages (0..100)</p>
                     </div>
                  </div>
                  <div class="topic nested1" id="gms-limitations"><a name="gms-limitations" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#gms-limitations" name="gms-limitations" shape="rect">Limitations</a></h3>
                     <div class="body">
                        <div class="p">
                           <ul class="ul">
                              <li class="li">
                                 <p class="p">If metric sets with NVLink are used but the links are not active,
                                    they may appear as fully utilized.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">Only one tool that subscribes to these counters can be used at a
                                    time, therefore, <span class="ph">Nsight Systems</span> GPU Metrics feature
                                    cannot be used at the same time as the following tools:
                                 </p>
                                 <ul class="ul">
                                    <li class="li">Nsight Graphics</li>
                                    <li class="li">Nsight Compute</li>
                                    <li class="li">DCGM (Data Center GPU Manager)
                                       
                                       <p class="p">Use the following command:</p>
                                       <ul class="ul">
                                          <li class="li">dcgmi profile --pause</li>
                                          <li class="li">dcgmi profile --resume</li>
                                       </ul>
                                       <p class="p">Or API:</p>
                                       <ul class="ul">
                                          <li class="li">dcgmProfPause</li>
                                          <li class="li">dcgmProfResume</li>
                                       </ul>
                                    </li>
                                    <li class="li">Non-NVIDIA products which use:
                                       
                                       <ul class="ul">
                                          <li class="li">CUPTI sampling used directly in the application. CUPTI trace is
                                             okay (although it will block <span class="ph">Nsight Systems</span> CUDA trace)
                                             
                                          </li>
                                          <li class="li">DCGM library</li>
                                       </ul>
                                    </li>
                                 </ul>
                              </li>
                              <li class="li"><span class="ph">Nsight Systems</span> limits the amount of memory that can be
                                 used to store GPU Metrics samples. Analysis with higher sampling rates
                                 or on GPUs with more SMs has a risk of exceeding this limit. This will
                                 lead to gaps on timeline filled with <samp class="ph codeph">Missing Data</samp>
                                 ranges. Future releases will reduce the frequency of this happening.
                                 
                              </li>
                           </ul>
                        </div>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="nvvideo-trace"><a name="nvvideo-trace" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#nvvideo-trace" name="nvvideo-trace" shape="rect">20.&nbsp;NVIDIA Video Codec SDK Trace</a></h2>
                  <div class="body">
                     <p class="p"><span class="ph">Nsight Systems</span> for x86 Linux and Windows targets can
                        trace calls from the NV Video Codec SDK. This software trace can be 
                        launched from the GUI or using the <samp class="ph codeph">--trace nvvideo</samp> from
                        the CLI
                     </p>
                     <p class="p"><img class="image" src="graphics/select-video-codec.png" alt="NV Video Codec SDK trace selection"></img></p>
                     <p class="p">On the timeline, calls on the CPU to the NV Encoder API and NV Decoder
                        API will be shown.
                     </p>
                     <p class="p"><img class="image" src="graphics/video-codec-trace.png" alt="NV Video Codec SDK trace in timeline"></img></p>
                  </div>
                  <div class="topic nested1" id="nvenc-default-function"><a name="nvenc-default-function" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#nvenc-default-function" name="nvenc-default-function" shape="rect">20.1.&nbsp;NV Encoder API Functions Traced by Default</a></h3>
                     <div class="body"><pre xml:space="preserve">
NvEncodeAPICreateInstance
nvEncOpenEncodeSession
nvEncGetEncodeGUIDCount
nvEncGetEncodeGUIDs
nvEncGetEncodeProfileGUIDCount
nvEncGetEncodeProfileGUIDs
nvEncGetInputFormatCount
nvEncGetInputFormats
nvEncGetEncodeCaps
nvEncGetEncodePresetCount
nvEncGetEncodePresetGUIDs
nvEncGetEncodePresetConfig
nvEncGetEncodePresetConfigEx
nvEncInitializeEncoder
nvEncCreateInputBuffer
nvEncDestroyInputBuffer
nvEncCreateBitstreamBuffer
nvEncDestroyBitstreamBuffer
nvEncEncodePicture
nvEncLockBitstream
nvEncUnlockBitstream
nvEncLockInputBuffer
nvEncUnlockInputBuffer
nvEncGetEncodeStats
nvEndGetSequenceParams
nvEncRegisterAsyncEvent
nvEncUnregisterAsyncEvent
nvEncMapInputResource
nvEncUnmapInputResource
nvEncDestroyEncoder
nvEncInvalidateRefFrames
nvEncOpenEncodeSessionEx
nvEncRegisterResource
nvEncUnregisterResource
nvEncReconfigureEncoder
nvEncCreateMVBuffer
nvEncDestroyMVBuffer
nvEncRunMotionEstimationOnly
nvEncGetLastErrorString
nvEncSetIOCudaStreams
nvEncGetSequenceParamEx
       </pre><p class="p"></p>
                        <p class="p hr"></p>
                        <p class="p"></p>
                     </div>
                  </div>
                  <div class="topic nested1" id="nvdec-default-function"><a name="nvdec-default-function" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#nvdec-default-function" name="nvdec-default-function" shape="rect">20.2.&nbsp;NV Decoder API Functions Traced by Default</a></h3>
                     <div class="body"><pre xml:space="preserve">
cuvidCreateVideoSource
cuvidCreateVideoSourceW
cuvidDestroyVideoSource
cuvidSetVideoSourceState
cudaVideoState
cuvidGetSourceVideoFormat
cuvidGetSourceAudioFormat
cuvidCreateVideoParser
cuvidParseVideoData
cuvidDestroyVideoParser
cuvidCreateDecoder
cuvidDestroyDecoder
cuvidDecodePicture
cuvidGetDecodeStatus
cuvidReconfigureDecoder
cuvidMapVideoFrame
cuvidUnmapVideoFrame
cuvidMapVideoFrame64
cuvidUnmapVideoFrame64
cuvidCtxLockCreate
cuvidCtxLockDestroy
cuvidCtxLock
cuvidCtxUnlock
       </pre><p class="p"></p>
                        <p class="p hr"></p>
                        <p class="p"></p>
                     </div>
                  </div>
                  <div class="topic nested1" id="nvjpeg-default-function"><a name="nvjpeg-default-function" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#nvjpeg-default-function" name="nvjpeg-default-function" shape="rect">20.3.&nbsp;NV JPEG API Functions Traced by Default</a></h3>
                     <div class="body"><pre xml:space="preserve">
nvjpegBufferDeviceCreate
nvjpegBufferDeviceDestroy
nvjpegBufferDeviceRetrieve
nvjpegBufferPinnedCreate
nvjpegBufferPinnedDestroy
nvjpegBufferPinnedRetrieve
nvjpegCreate
nvjpegCreateEx
nvjpegCreateSimple
nvjpegDecode
nvjpegDecodeBatched
nvjpegDecodeBatchedEx
nvjpegDecodeBatchedInitialize
nvjpegDecodeBatchedPreAllocate
nvjpegDecodeBatchedSupported
nvjpegDecodeBatchedSupportedEx
nvjpegDecodeJpeg
nvjpegDecodeJpegDevice
nvjpegDecodeJpegHost
nvjpegDecodeJpegTransferToDevice
nvjpegDecodeParamsCreate
nvjpegDecodeParamsDestroy
nvjpegDecodeParamsSetAllowCMYK
nvjpegDecodeParamsSetOutputFormat
nvjpegDecodeParamsSetROI
nvjpegDecodeParamsSetScaleFactor
nvjpegDecoderCreate
nvjpegDecoderDestroy
nvjpegDecoderJpegSupported
nvjpegDecoderStateCreate
nvjpegDestroy
nvjpegEncodeGetBufferSize
nvjpegEncodeImage
nvjpegEncodeRetrieveBitstream
nvjpegEncodeRetrieveBitstreamDevice
nvjpegEncoderParamsCopyHuffmanTables
nvjpegEncoderParamsCopyMetadata
nvjpegEncoderParamsCopyQuantizationTables
nvjpegEncoderParamsCreate
nvjpegEncoderParamsDestroy
nvjpegEncoderParamsSetEncoding
nvjpegEncoderParamsSetOptimizedHuffman
nvjpegEncoderParamsSetQuality
nvjpegEncoderParamsSetSamplingFactors
nvjpegEncoderStateCreate
nvjpegEncoderStateDestroy
nvjpegEncodeYUV,(nvjpegHandle_t handle
nvjpegGetCudartProperty
nvjpegGetDeviceMemoryPadding
nvjpegGetImageInfo
nvjpegGetPinnedMemoryPadding
nvjpegGetProperty
nvjpegJpegStateCreate
nvjpegJpegStateDestroy
nvjpegJpegStreamCreate
nvjpegJpegStreamDestroy
nvjpegJpegStreamGetChromaSubsampling
nvjpegJpegStreamGetComponentDimensions
nvjpegJpegStreamGetComponentsNum
nvjpegJpegStreamGetFrameDimensions
nvjpegJpegStreamGetJpegEncoding
nvjpegJpegStreamParse
nvjpegJpegStreamParseHeader
nvjpegSetDeviceMemoryPadding
nvjpegSetPinnedMemoryPadding
nvjpegStateAttachDeviceBuffer
nvjpegStateAttachPinnedBuffer
       </pre></div>
                  </div>
               </div>
               <div class="topic nested0" id="communication-trace"><a name="communication-trace" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#communication-trace" name="communication-trace" shape="rect">21.&nbsp;Network Communication Profiling</a></h2>
                  <div class="body">
                     <p class="p"></p>
                     <p class="p"><span class="ph">Nsight Systems</span> can be used to profiles several popular
                        network communication protocols. To enable this, please select the
                        <strong class="ph b">Communication profiling options</strong> dropdown. 
                     </p>
                     <p class="p"><img class="image" src="graphics/project-settings-all.png" alt="Project settings screen"></img></p>
                     <p class="p">Then select the libraries you would like to trace:</p>
                     <p class="p"><img class="image" src="graphics/project-settings-communication.png" alt="Communication library selection screen"></img></p>
                  </div>
                  <div class="topic nested1" id="mpi-trace"><a name="mpi-trace" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#mpi-trace" name="mpi-trace" shape="rect">21.1.&nbsp;MPI API Trace</a></h3>
                     <div class="body">
                        <p class="p">For Linux x86_64, ARM and Power targets, <span class="ph">Nsight Systems</span> is
                           capable of capturing information about the MPI APIs executed in the profiled
                           process. It has built-in API trace support for Open MPI and MPICH based MPI
                           implementations.
                        </p>
                        <p class="p"><img class="image" src="graphics/mpi-project.png" alt="MPI trace selection"></img></p>
                        <p class="p">Only a subset of the MPI API, including blocking and non-blocking
                           point-to-point and collective communication, and file I/O operations, is traced.
                           If you require more control over the list of traced APIs or if you are using
                           a different MPI implementation, you can use the
                           <a class="xref" href="https://github.com/NVIDIA/cuda-profiler/tree/master/nvtx_pmpi_wrappers" target="_blank" shape="rect">NVTX wrappers for MPI</a>. If you set the
                           environment variable <samp class="ph codeph">LD_PRELOAD</samp> to the path of generated
                           wrapper library, <span class="ph">Nsight Systems</span> will capture and report the MPI
                           API trace information when NVTX tracing is enabled. Choose an NVTX domain
                           name other than "MPI", since it is filtered out by <span class="ph">Nsight Systems</span>
                           when MPI tracing is not enabled.
                        </p>
                        <p class="p"><img class="image" src="graphics/mpi-trace.png" alt="MPI API trace"></img></p>
                        <p class="p"><strong class="ph b">MPI Communication Parameters</strong></p>
                        <p class="p"><span class="ph">Nsight Systems</span> can get additional information about MPI
                           communication parameters. Currently, the parameters are only visible in the
                           mouseover tooltips or in the eventlog. This means that the data is only
                           available via the GUI. Future versions of the tool will export this
                           information into the SQLite data files for postrun analysis.
                        </p>
                        <p class="p">In order to fully interpret MPI communications, data for all ranks
                           associated with a communication operation must be loaded into
                           <span class="ph">Nsight Systems</span>.
                        </p>
                        <p class="p">Here is an example of MPI_COMM_WORLD data. This does not require any
                           additional team data, since local rank is the same as global rank.
                        </p>
                        <p class="p">(Screenshot shows communication parameters for an MPI_Bcast call on rank
                           3)
                        </p>
                        <p class="p"><img class="image" src="graphics/mpi-comparam1.png" alt="MPI communication parameter trace"></img></p>
                        <p class="p">When not all processes that are involved in an MPI communication are
                           loaded into <span class="ph">Nsight Systems</span> the following information is
                           available.
                        </p>
                        <ul class="ul">
                           <li class="li">Right-hand screenshot shows a reused communicator handle (last number
                              increased).
                           </li>
                           <li class="li">Encoding: MPI_COMM[*team size*]*global-group-root-rank*.*group-ID*</li>
                        </ul>
                        <p class="p"><img class="image" src="graphics/mpi-comparam2.png" alt="MPI communication parameter trace"></img></p>
                        <p class="p"> When all reports are loaded into <span class="ph">Nsight Systems</span>:
                        </p>
                        <ul class="ul">
                           <li class="li">World rank is shown in addition to group-local rank "(world rank X)"</li>
                           <li class="li">Encoding: MPI_COMM[*team size*]{rank0, rank1, ...}</li>
                           <li class="li">At most 8 ranks are shown (the numbers represent world ranks, the
                              position in the list is the group-local rank)
                           </li>
                        </ul>
                        <p class="p"><img class="image" src="graphics/mpi-comparam3.png" alt="MPI communication parameter trace"></img></p>
                        <p class="p"><strong class="ph b">MPI functions traced:</strong></p><pre xml:space="preserve">
MPI_Init[_thread], MPI_Finalize
MPI_Send, MPI_{B,S,R}send, MPI_Recv, MPI_Mrecv
MPI_Sendrecv[_replace]

MPI_Barrier, MPI_Bcast
MPI_Scatter[v], MPI_Gather[v]
MPI_Allgather[v], MPI_Alltoall[{v,w}]
MPI_Allreduce, MPI_Reduce[_{scatter,scatter_block,local}]
MPI_Scan, MPI_Exscan

MPI_Isend, MPI_I{b,s,r}send, MPI_I[m]recv
MPI_{Send,Bsend,Ssend,Rsend,Recv}_init
MPI_Start[all]
MPI_Ibarrier, MPI_Ibcast
MPI_Iscatter[v], MPI_Igather[v]
MPI_Iallgather[v], MPI_Ialltoall[{v,w}]
MPI_Iallreduce, MPI_Ireduce[{scatter,scatter_block}]
MPI_I[ex]scan
MPI_Wait[{all,any,some}]

MPI_File_{open,close,delete,sync}
MPI_File_{read,write}[_{all,all_begin,all_end}]
MPI_File_{read,write}_at[_{all,all_begin,all_end}]
MPI_File_{read,write}_shared
MPI_File_{read,write}_ordered[_{begin,end}]
MPI_File_i{read,write}[_{all,at,at_all,shared}]
MPI_File_set_{size,view,info}
MPI_File_get_{size,view,info,group,amode}
MPI_File_preallocate</pre><p class="p"></p>
                        <p class="p hr"></p>
                        <p class="p"></p>
                     </div>
                  </div>
                  <div class="topic nested1" id="openshmem-trace"><a name="openshmem-trace" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#openshmem-trace" name="openshmem-trace" shape="rect">21.2.&nbsp;OpenSHMEM Library Trace</a></h3>
                     <div class="body">
                        <p class="p">If OpenSHMEM library trace is selected <span class="ph">Nsight Systems</span>
                           will trace the subset of OpenSHMEM API functions that are most likely be
                           involved in performance bottlenecks. To keep overhead low
                           <span class="ph">Nsight Systems</span> does not trace all functions.
                        </p>
                        <p class="p"><strong class="ph b">OpenSHMEM 1.5 Functions Not Traced</strong></p><pre xml:space="preserve">
shmem_my_pe
shmem_n_pes
shmem_global_exit
shmem_pe_accessible
shmem_addr_accessible
shmem_ctx_{create,destroy,get_team}
shmem_global_exit
shmem_info_get_{version,name}
shmem_{my_pe,n_pes,pe_accessible,ptr}
shmem_query_thread
shmem_team_{create_ctx,destroy}
shmem_team_get_config
shmem_team_{my_pe,n_pes,translate_pe}
shmem_team_split_{2d,strided}
shmem_test*
</pre></div>
                  </div>
                  <div class="topic nested1" id="ucx-trace"><a name="ucx-trace" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#ucx-trace" name="ucx-trace" shape="rect">21.3.&nbsp;UCX Library Trace</a></h3>
                     <div class="body">
                        <p class="p">If UCX library trace is selected <span class="ph">Nsight Systems</span>
                           will trace the subset of functions of the UCX protocol layer UCP that are
                           most likely be involved in performance bottlenecks. To keep overhead low
                           <span class="ph">Nsight Systems</span> does not trace all functions.
                        </p>
                        <p class="p"><strong class="ph b">UCX functions traced:</strong></p><pre xml:space="preserve">
ucp_am_send_nb[x]
ucp_am_recv_data_nbx
ucp_am_data_release
ucp_atomic_{add{32,64},cswap{32,64},fadd{32,64},swap{32,64}}
ucp_atomic_{post,fetch_nb,op_nbx}
ucp_cleanup
ucp_config_{modify,read,release}
ucp_disconnect_nb
ucp_dt_{create_generic,destroy}
ucp_ep_{create,destroy,modify_nb,close_nbx}
ucp_ep_flush[{_nb,_nbx}]
ucp_listener_{create,destroy,query,reject}
ucp_mem_{advise,map,unmap,query}
ucp_{put,get}[_nbi]
ucp_{put,get}_nb[x]
ucp_request_{alloc,cancel,is_completed}
ucp_rkey_{buffer_release,destroy,pack,ptr}
ucp_stream_data_release
ucp_stream_recv_data_nb
ucp_stream_{send,recv}_nb[x]
ucp_stream_worker_poll
ucp_tag_msg_recv_nb[x]
ucp_tag_{send,recv}_nbr
ucp_tag_{send,recv}_nb[x]
ucp_tag_send_sync_nb[x]
ucp_worker_{create,destroy,get_address,get_efd,arm,fence,wait,signal,wait_mem}
ucp_worker_flush[{_nb,_nbx}]
ucp_worker_set_am_{handler,recv_handler}
      </pre><p class="p"><strong class="ph b">UCX Functions Not Traced:</strong></p><pre xml:space="preserve">
ucp_config_print
ucp_conn_request_query
ucp_context_{query,print_info}
ucp_get_version[_string]
ucp_ep_{close_nb,print_info,query,rkey_unpack}
ucp_mem_print_info
ucp_request_{check_status,free,query,release,test}
ucp_stream_recv_request_test
ucp_tag_probe_nb
ucp_tag_recv_request_test
ucp_worker_{address_query,print_info,progress,query,release_address}
       </pre><p class="p">Additional API functions from other UCX layers may be added in a future
                           version of the product.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="nv-communications"><a name="nv-communications" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#nv-communications" name="nv-communications" shape="rect">21.4.&nbsp;NVIDIA NVSHMEM and NCCL Trace</a></h3>
                     <div class="body">
                        <p class="p">The NVIDIA network communication libraries NVSHMEM and NCCL have been
                           instrumented using NVTX annotations. To enable tracing these libraries in
                           <span class="ph">Nsight Systems</span>, turn on NVTX tracing in the GUI or
                           CLI. To enable the NVTX instrumentation of the NVSHMEM library, make sure
                           that the environment variable <samp class="ph codeph">NVSHMEM_NVTX</samp> is set properly,
                           e.g. <samp class="ph codeph">NVSHMEM_NVTX=common</samp>.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="NIC-sampling"><a name="NIC-sampling" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#NIC-sampling" name="NIC-sampling" shape="rect">21.5.&nbsp;NIC Metric Sampling</a></h3>
                     <div class="body">
                        <p class="p">This feature is experimental.</p>
                        <p class="p"><strong class="ph b">Overview</strong></p>
                        <p class="p">NVIDIA ConnectX smart network interface cards (smart NICs) offer
                           advanced hardware offloads and accelerations for network operations.
                           Viewing smart NICs metrics, on <span class="ph">Nsight Systems</span> timeline,
                           enables developers to better understand their application’s network usage.
                           Developers can use this information to optimize the application’s
                           performance.
                        </p>
                        <p class="p"><strong class="ph b">Limitations/Requirements</strong></p>
                        <ul class="ul">
                           <li class="li">NIC metric sampling supports NVIDIA ConnectX boards
                              starting with ConnectX 5
                           </li>
                           <li class="li">NIC metric sampling supported on Linux x86_64 machines only, having
                              minimum Linux kernel 4.12 and minimum
                              MLNX_OFED 4.1.
                           </li>
                           <li class="li">NIC metric sampling is only available from
                              the command line
                           </li>
                        </ul>
                        <p class="p"><strong class="ph b">Collecting NIC Metrics Using the Command Line</strong></p>
                        <p class="p">To collect NIC performance metric, using <span class="ph">Nsight Systems</span>
                           CLI, add the <samp class="ph codeph">--nic-metrics</samp> command line switch:
                        </p><pre xml:space="preserve">nsys profile --nic-metrics=true my_app</pre><p class="p"><img class="image" src="graphics/nic-metrics-sampling.png" alt="NIC metric sampling screenshot"></img></p>
                        <p class="p"><strong class="ph b">Available Metrics</strong></p>
                        <ul class="ul">
                           <li class="li"><strong class="ph b">Bytes sent</strong> - Number of bytes sent through all NIC ports.
                           </li>
                           <li class="li"><strong class="ph b">Packets sent</strong> - Number of packets sent through all NIC ports.
                           </li>
                           <li class="li"><strong class="ph b">Bytes received</strong> - Number of bytes received by all NIC ports.
                           </li>
                           <li class="li"><strong class="ph b">Packets received</strong>  - Number of packets received by all NIC
                              ports.
                           </li>
                           <li class="li"><strong class="ph b">CNPs sent</strong> - Number of congestion notification packets sent by the
                              NIC.
                           </li>
                           <li class="li"><strong class="ph b">CNPs received</strong> - Number of congestion notification packets
                              received and handled by the NIC.
                           </li>
                           <li class="li"><strong class="ph b">Send waits</strong> - The number of ticks during which ports had data to
                              transmit but no data was sent during the entire tick (either because of
                              insufficient credits or because of lack of arbitration)
                           </li>
                        </ul>
                        <p class="p"><strong class="ph b">Usage Examples</strong></p>
                        <ul class="ul">
                           <li class="li">The <samp class="ph codeph">Bytes sent/sec</samp> and the <samp class="ph codeph">Bytes
                                 received/sec</samp> metrics enables identifying idle and busy NIC times.
                              
                              <ul class="ul">
                                 <li class="li">Developers may shift network operations from busy to idle times
                                    to reduce network congestion and latency.
                                 </li>
                                 <li class="li">Developers can use idle NIC times to send additional data without
                                    reducing application performance.
                                 </li>
                              </ul>
                           </li>
                           <li class="li">CNPs (congestion notification packets) received/sent and Send waits
                              metrics may explain network latencies. A developer seeing the time periods
                              when the network was congested may rewrite his algorithm to avoid the
                              observed congestions.
                           </li>
                           <li class="li">Comparing the ratio between bytes sent and packets sent can give
                              developers a hint about the average packet size their application is using.
                              The larger the packet size, the higher the practical bandwidth.
                           </li>
                        </ul>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="report"><a name="report" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#report" name="report" shape="rect">22.&nbsp;Reading Your Report in GUI</a></h2>
                  <div class="body">
                     <p class="p"></p>
                  </div>
                  <div class="topic nested1" id="generate-new-report"><a name="generate-new-report" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#generate-new-report" name="generate-new-report" shape="rect">22.1.&nbsp;Generating a New Report</a></h3>
                     <div class="body">
                        <p class="p">
                           Users can generate a new report by stopping a profiling session.  If a
                           profiling session has been canceled, a report will not be generated, and
                           all collected data will be discarded.
                        </p>
                        <p class="p">
                           A new <samp class="ph codeph">.nsys-rep</samp> file will be created and put into the same
                           directory as the project file (<samp class="ph codeph">.qdproj</samp>).
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="open-report"><a name="open-report" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#open-report" name="open-report" shape="rect">22.2.&nbsp;Opening an Existing Report</a></h3>
                     <div class="body">
                        <p class="p">
                           An existing <samp class="ph codeph">.nsys-rep</samp> file can be opened using <strong class="ph b">File &gt;
                              Open...</strong>.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="report-sharing"><a name="report-sharing" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#report-sharing" name="report-sharing" shape="rect">22.3.&nbsp;Sharing a Report File</a></h3>
                     <div class="body">
                        <p class="p">
                           Report files (<samp class="ph codeph">.nsys-rep</samp>) are self-contained and can be
                           shared with other users of <span class="ph">Nsight Systems</span>.  The only
                           requirement is that the same or newer version of
                           <span class="ph">Nsight Systems</span> is always used to open report files.
                        </p>
                        <p class="p">
                           Project files (<samp class="ph codeph">.qdproj</samp>) are currently not shareable,
                           since they contain full paths to the report files.
                        </p>
                        <p class="p">
                           To quickly navigate to the directory containing the report file, right
                           click on it in the Project Explorer, and choose <strong class="ph b">Show in folder...</strong>
                           in the context menu.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="report-tab"><a name="report-tab" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#report-tab" name="report-tab" shape="rect">22.4.&nbsp;Report Tab</a></h3>
                     <div class="body">
                        <p class="p">
                           While generating a new report or loading an existing one, a new tab will
                           be created.  The most important parts of the report tab are:
                        </p>
                        <ul class="ul">
                           <li class="li">
                              <p class="p"><strong class="ph b">View selector</strong> —  Allows switching between <em class="ph i">Analysis Summary</em>,
                                 <em class="ph i">Timeline View</em>, <em class="ph i">Diagnostics Summary</em>, and <em class="ph i">Symbol
                                    Resolution Logs</em> views.
                              </p>
                              <blockquote class="lq">
                                 <p class="p"><img class="image" src="graphics/report-view-selector.png" alt="Page selector"></img></p>
                              </blockquote>
                           </li>
                           <li class="li">
                              <p class="p"><strong class="ph b">Timeline</strong> —  This is where all charts are displayed.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p"><strong class="ph b">Function table</strong> —  Located below the timeline, it displays
                                 statistical information about functions in the target application in
                                 multiple ways.
                              </p>
                           </li>
                        </ul>
                        <p class="p">Additionally, the following controls are available:</p>
                        <ul class="ul">
                           <li class="li">
                              <p class="p"><strong class="ph b">Zoom slider</strong> —  Allows you to vertically zoom the charts on the
                                 timeline.
                              </p>
                           </li>
                        </ul>
                     </div>
                  </div>
                  <div class="topic nested1" id="analysis-summary-view"><a name="analysis-summary-view" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#analysis-summary-view" name="analysis-summary-view" shape="rect">22.5.&nbsp;Analysis Summary View</a></h3>
                     <div class="body">
                        <p class="p">
                           This view shows a summary of the profiling session. In particular, it is
                           useful to review the project configuration used to generate this report.
                           Information from this view can be selected and copied using the mouse
                           cursor.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="timeline-view"><a name="timeline-view" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#timeline-view" name="timeline-view" shape="rect">22.6.&nbsp;Timeline View</a></h3>
                     <div class="body">
                        <p class="p">
                           The timeline view consists of two main controls: the timeline at the top,
                           and a bottom pane that contains the events view and the function table.
                           In some cases, when sampling of a process has not been enabled, the
                           function table might be empty and hidden.
                        </p>
                        <p class="p">
                           The bottom view selector sets the view that is displayed in the bottom
                           pane.
                        </p>
                        <blockquote class="lq">
                           <p class="p"><img class="image" src="graphics/bottom-view-selector.png" alt="Bottom view selection"></img></p>
                        </blockquote>
                     </div>
                     <div class="topic nested2" id="timeline"><a name="timeline" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#timeline" name="timeline" shape="rect">22.6.1.&nbsp;Timeline</a></h3>
                        <div class="body">
                           <p class="p">
                              Timeline is a versatile control that contains a tree-like <strong class="ph b">hierarchy</strong>
                              on the left, and corresponding <em class="ph i">charts</em> on the right.
                           </p>
                           <p class="p">
                              Contents of the hierarchy depend on the project settings used to collect
                              the report. For example, if a certain feature has not been enabled,
                              corresponding rows will not be show on the timeline.
                           </p>
                           <div class="p">
                              To generate a timeline screenshot without opening the full GUI, use the
                              command <pre xml:space="preserve">nsys-ui.exe --screenshot filename.nsys-rep</pre></div>
                           <p class="p">
                              To display trace events in the Events View right-click a timeline row and
                              select the “Show in Events View” command. The events of the selected row
                              and all of its sub-rows will be displayed in the Events View.
                           </p>
                           <p class="p">
                              If a timeline row has been selected for display in the Events View then
                              double-clicking a timeline item on that row will automatically scroll the
                              content of the Events View to make the corresponding Events View item
                              visible and select it.
                           </p>
                        </div>
                        <div class="topic nested3" id="rowheight"><a name="rowheight" shape="rect">
                              <!-- --></a><h3 class="title topictitle2"><a href="#rowheight" name="rowheight" shape="rect">Row Height</a></h3>
                           <div class="body">
                              <p class="p">
                                 		  Several of the rows in the timeline use height as a way to model
                                 		  the percent utilization of resources. This gives the user insight into
                                 		  what is going on even when the timeline is zoomed all the way out.
                              </p>
                              <p class="p"><img class="image" src="graphics/row-height.png" alt="CUDA row heights"></img></p>
                              <p class="p">
                                 		  In this picture you see that for kernel occupation there is a colored
                                 		  bar of variable height.
                              </p>
                              <p class="p"><span class="ph">Nsight Systems</span> calculates the average occupancy for the
                                 		  period of time represented by particular pixel width of screen. It then
                                 		  uses that average to set the top of the colored section. So, for
                                 		  instance, if 25% of that timeslice the kernel is active, the bar goes
                                 		  25% of the distance to the top of the row.
                              </p>
                              <p class="p">
                                 		  In order to make the difference clear, if the percentage of the row
                                 		  height is non-zero, but would be represented by less than one vertical
                                 		  pixel, <span class="ph">Nsight Systems</span> displays it as one pixel high.
                                 		  The gray height represents the maximum usage in that time range.
                              </p>
                              <p class="p">
                                 		  This row height coding is used in the CPU utilization, thread and
                                 		  process occupancy, kernel occupancy, and memory transfer activity rows.
                                 		  
                              </p>
                           </div>
                        </div>
                     </div>
                     <div class="topic nested2" id="events-view"><a name="events-view" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#events-view" name="events-view" shape="rect">22.6.2.&nbsp;Events View</a></h3>
                        <div class="body">
                           <p class="p">
                              The Events View provides a tabular display of the trace events. The view
                              contents can be searched and sorted.
                           </p>
                           <p class="p">
                              Double-clicking an item in the Events View automatically focuses the
                              Timeline View on the corresponding timeline item.
                           </p>
                           <p class="p">
                              API calls, GPU executions, and debug markers that occurred within the
                              boundaries of a debug marker are displayed nested to that debug marker.
                              Multiple levels of nesting are supported.
                           </p>
                           <p class="p">Events view recognizes these types of debug markers:</p>
                           <ul class="ul">
                              <li class="li">
                                 <p class="p">NVTX</p>
                              </li>
                              <li class="li">
                                 <p class="p">Vulkan VK_EXT_debug_marker markers, VK_EXT_debug_utils labels</p>
                              </li>
                              <li class="li">
                                 <p class="p">PIX events and markers</p>
                              </li>
                              <li class="li">
                                 <p class="p">OpenGL KHR_debug markers</p>
                              </li>
                           </ul>
                           <p class="p"><img class="image" src="graphics/events-view-nested-markers.png" alt="Events View nested debug markers"></img></p>
                           <p class="p">You can copy and paste from the events view by highlighting rows, using
                              <strong class="ph b">Shift</strong> or <strong class="ph b">Ctrl</strong> to enable multi-select. Right clicking on the
                              selection will give you a copy option. 
                           </p>
                           <p class="p"><img class="image" src="graphics/event-copy.png" alt="Events View copy selection"></img></p>
                           <p class="p">Pasting into text gives you a tab separated view:</p>
                           <p class="p"><img class="image" src="graphics/event-paste-notepad.png" alt="Events View paste into notepad"></img></p>
                           <p class="p">Pasting into spreadsheet properly copies into rows and columns:</p>
                           <p class="p"><img class="image" src="graphics/event-paste-spreadsheet.png" alt="Events View paste into spreadsheet"></img></p>
                        </div>
                     </div>
                     <div class="topic nested2" id="function-table-modes"><a name="function-table-modes" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#function-table-modes" name="function-table-modes" shape="rect">22.6.3.&nbsp;Function Table Modes</a></h3>
                        <div class="body">
                           <blockquote class="lq">
                              <p class="p"><img class="image" src="graphics/function-table-modes.png" alt="Function table modes"></img></p>
                           </blockquote>
                           <p class="p">The function table can work in three modes:</p>
                           <ul class="ul">
                              <li class="li">
                                 <p class="p"><strong class="ph b">Top-Down View</strong> — In this mode, expanding top-level functions
                                    provides information about the <em class="ph i">callee</em> functions. One of the
                                    top-level functions is typically the main function of your application,
                                    or another entry point defined by the runtime libraries.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><strong class="ph b">Bottom-Up View</strong> — This is a reverse of the Top-Down view. On the
                                    top level, there are functions directly hit by the sampling profiler. To
                                    explore all possible call chains leading to these functions, you need to
                                    expand the subtrees of the top-level functions.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><strong class="ph b">Flat View</strong> — This view enumerates all functions ever observed by
                                    the profiler, even if they have never been directly hit, but just
                                    appeared somewhere on the call stack. This view typically provides a
                                    high-level overview of which parts of the code are CPU-intensive.
                                 </p>
                              </li>
                           </ul>
                           <p class="p">
                              Each of the views helps understand particular performance issues of the
                              application being profiled.  For example:
                           </p>
                           <ul class="ul">
                              <li class="li">
                                 <p class="p">When trying to find specific bottleneck functions that can be optimized,
                                    the Bottom-Up view should be used.  Typically, the top few functions
                                    should be examined. Expand them to understand in which contexts they are
                                    being used.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">To navigate the call tree of the application and while generally
                                    searching for algorithms and parts of the code that consume unexpectedly
                                    large amount of CPU time, the Top-Down view should be used.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">To quickly assess which parts of the application, or high level parts of
                                    an algorithm, consume significant amount of CPU time, use the Flat view.
                                 </p>
                              </li>
                           </ul>
                           <p class="p">
                              The Top-Down and Bottom-Up views have <em class="ph i">Self</em> and <em class="ph i">Total</em>
                              columns, while the Flat view has a <em class="ph i">Flat</em> column. It is important to
                              understand the meaning of each of the columns:
                           </p>
                           <ul class="ul">
                              <li class="li">
                                 <p class="p">Top-Down view</p>
                                 <ul class="ul">
                                    <li class="li">
                                       <p class="p"><strong class="ph b">Self</strong> column denotes the relative amount of time spent executing
                                          instructions of this particular function.
                                       </p>
                                    </li>
                                    <li class="li">
                                       <p class="p"><strong class="ph b">Total</strong> column shows how much time has been spent executing this
                                          function, including all other functions called from this one. Total
                                          values of sibling rows sum up to the Total value of the parent row, or
                                          100% for the top-level rows.
                                       </p>
                                    </li>
                                 </ul>
                              </li>
                              <li class="li">
                                 <p class="p">Bottom-Up view</p>
                                 <ul class="ul">
                                    <li class="li">
                                       <p class="p"><strong class="ph b">Self</strong> column for <em class="ph i">top-level rows</em>, as in the Top-Down view,
                                          shows how much time has been spent directly in this function. Self
                                          times of all top-level rows add up to 100%.
                                       </p>
                                    </li>
                                    <li class="li">
                                       <p class="p"><strong class="ph b">Self</strong> column for <em class="ph i">children rows</em> breaks down the value of
                                          the parent row based on the various call chains leading to that
                                          function. Self times of sibling rows add up to the value of the parent
                                          row.
                                       </p>
                                    </li>
                                 </ul>
                              </li>
                              <li class="li">
                                 <p class="p">Flat view</p>
                                 <ul class="ul">
                                    <li class="li">
                                       <p class="p"><strong class="ph b">Flat</strong> column shows how much time this function has been anywhere
                                          on the call stack.  Values in this column do not add up or have other
                                          significant relationships.
                                       </p>
                                    </li>
                                 </ul>
                                 <blockquote class="lq">
                                    <div class="p">
                                       <div class="tablenoborder">
                                          <table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
                                             <tbody class="tbody">
                                                <tr class="row">
                                                   <td class="entry nvidiagreen" align="center" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                                      <p class="p">&nbsp;&nbsp;<strong class="ph b">Note:</strong>&nbsp;&nbsp;
                                                      </p>
                                                   </td>
                                                   <td class="entry notebox" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                                      <p class="p">If low-impact functions have been filtered out, values may
                                                         not add up correctly to 100%, or to the value of the parent row. This
                                                         filtering can be disabled.
                                                      </p>
                                                   </td>
                                                </tr>
                                             </tbody>
                                          </table>
                                       </div>
                                    </div>
                                 </blockquote>
                                 <p class="p">Contents of the symbols table is tightly related to the timeline. Users
                                    can apply and modify filters on the timeline, and they will affect
                                    which information is displayed in the symbols table:
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><strong class="ph b">Per-thread filtering</strong> — Each thread that has sampling information
                                    associated with it has a checkbox next to it on the timeline. Only
                                    threads with selected checkboxes are represented in the symbols table.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><strong class="ph b">Time filtering</strong> — A time filter can be setup on the timeline by
                                    pressing the left mouse button, dragging over a region of interest on the
                                    timeline, and then choosing <strong class="ph b">Filter by selection</strong> in the dropdown
                                    menu. In this case, only sampling information collected during the
                                    selected time range will be used to build the symbols table.
                                 </p>
                                 <div class="p">
                                    <div class="tablenoborder">
                                       <table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
                                          <tbody class="tbody">
                                             <tr class="row">
                                                <td class="entry nvidiagreen" align="center" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                                   <p class="p">&nbsp;&nbsp;<strong class="ph b">Note:</strong>&nbsp;&nbsp;
                                                   </p>
                                                </td>
                                                <td class="entry notebox" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                                   <p class="p">If too little sampling data is being used to build the symbols
                                                      table (for example, when the sampling rate is configured to be low, and a
                                                      short period of time is used for time-based filtering), the numbers in
                                                      the symbols table might not be representative or accurate in some cases.
                                                      
                                                   </p>
                                                </td>
                                             </tr>
                                          </tbody>
                                       </table>
                                    </div>
                                 </div>
                              </li>
                           </ul>
                        </div>
                     </div>
                     <div class="topic nested2" id="filter-dialog"><a name="filter-dialog" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#filter-dialog" name="filter-dialog" shape="rect">22.6.4.&nbsp;Filter Dialog</a></h3>
                        <div class="body">
                           <blockquote class="lq">
                              <p class="p"><img class="image" src="graphics/report-filter-dialog.png" alt="Filter dialog"></img></p>
                           </blockquote>
                           <ul class="ul">
                              <li class="li"><strong class="ph b">Collapse unresolved lines</strong> is useful if some of the binary code does
                                 not have symbols. In this case, subtrees that consist of only unresolved
                                 symbols get collapsed in the Top-Down view, since they provide very
                                 little useful information.
                              </li>
                              <li class="li"><strong class="ph b">Hide functions with CPU usage below X%</strong> is useful for large
                                 applications, where the sampling profiler hits lots of function just a
                                 few times. To filter out the "long tail," which is typically not
                                 important for CPU performance bottleneck analysis, this checkbox should
                                 be selected.
                              </li>
                           </ul>
                        </div>
                     </div>
                  </div>
                  <div class="topic nested1" id="diagnostics-view"><a name="diagnostics-view" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#diagnostics-view" name="diagnostics-view" shape="rect">22.7.&nbsp;Diagnostics Summary View</a></h3>
                     <div class="body">
                        <p class="p">
                           This view shows important messages. Some of them were generated during
                           the profiling session, while some were added while processing and
                           analyzing data in the report.  Messages can be one of the following
                           types:
                        </p>
                        <ul class="ul">
                           <li class="li">
                              <p class="p">Informational messages</p>
                           </li>
                           <li class="li">
                              <p class="p">Warnings</p>
                           </li>
                           <li class="li">
                              <p class="p">Errors</p>
                           </li>
                        </ul>
                        <p class="p">
                           To draw attention to important diagnostics messages, a summary line is
                           displayed on the timeline view in the top right corner:
                        </p>
                        <blockquote class="lq">
                           <p class="p"><img class="image" src="graphics/report-diagnostics-link.png" alt="Diagnostics messages"></img></p>
                        </blockquote>
                        <p class="p">
                           Information from this view can be selected and copied using the mouse
                           cursor.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="symbol-resolution-view"><a name="symbol-resolution-view" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#symbol-resolution-view" name="symbol-resolution-view" shape="rect">22.8.&nbsp;Symbol Resolution Logs View</a></h3>
                     <div class="body">
                        <p class="p">
                           This view shows all messages related to the process of resolving symbols.
                           It might be useful to debug issues when some of the symbol names in the
                           symbols table of the timeline view are unresolved.
                        </p>
                        <p class="p"></p>
                        <p class="p hr"></p>
                        <p class="p"></p>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="multi-report-support"><a name="multi-report-support" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#multi-report-support" name="multi-report-support" shape="rect">23.&nbsp;Adding Report to the Timeline</a></h2>
                  <div class="body">
                     <p class="p">
                        Starting with 2021.3, <span class="ph">Nsight Systems</span> can load multiple
                        report files into a single timeline.
                        This is a BETA feature and will be improved in the future releases.
                        Please let us know about your experience on the forums or through
                        <strong class="ph b">Help &gt; Send Feedback...</strong> in the main menu.
                     </p>
                     <p class="p">
                        To load multiple report files into a single timeline, first start by
                        opening a report as usual — using <strong class="ph b">File &gt; Open...</strong> from the main menu,
                        or double clicking on a report in the Project Explorer window. Then additional
                        report files can be loaded into the same timeline using one of the methods:
                     </p>
                     <ul class="ul">
                        <li class="li">
                           <p class="p"><strong class="ph b">File &gt; Add Report (beta)...</strong> in the main menu,
                              and select another report file that you want to open
                           </p>
                        </li>
                        <li class="li">
                           <p class="p">Right click on the report in the project explorer window, and click
                              <strong class="ph b">Add Report (beta)</strong></p>
                        </li>
                     </ul>
                     <p class="p"><img class="image" src="graphics/report-add-another.png" alt="TODO"></img></p>
                  </div>
                  <div class="topic nested1" id="multi-report-time-sync"><a name="multi-report-time-sync" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#multi-report-time-sync" name="multi-report-time-sync" shape="rect">23.1.&nbsp;Time Synchronization</a></h3>
                     <div class="body">
                        <p class="p">When multiple reports are loaded into a single timeline, timestamps between them need
                           to be adjusted, such that events that happened at the same time appear to be aligned.
                        </p>
                        <p class="p"><span class="ph">Nsight Systems</span> can automatically adjust timestamps based on <strong class="ph b">UTC time</strong>
                           recorded around the collection start time. This method is used by default when other
                           more precise methods are not available. This time can be seen as
                           <samp class="ph codeph">UTC time at t=0</samp> in the <em class="ph i">Analysis Summary</em> page of the report file.
                           Refer to your OS documentation to learn how to sync the software clock using the
                           Network Time Protocol (NTP). NTP-based time synchronization is not very precise,
                           with the typical errors on the scale of one to tens of milliseconds.
                        </p>
                        <p class="p">Reports collected on the same physical machine can use synchronization based on
                           <strong class="ph b">Timestamp Counter (TSC) values</strong>. These are platform-specific counters,
                           typically accessed in user space applications using the RDTSC instruction on
                           x86_64 architecture, or by reading the CNTVCT register on Arm64. Their values
                           converted to nanoseconds can be seen as <samp class="ph codeph">TSC value at t=0</samp>
                           in the <em class="ph i">Analysis Summary</em> page of the report file. Reports synchronized using TSC
                           values can be aligned with nanoseconds-level precision.
                        </p>
                        <p class="p">TSC-based time synchronization is activated automatically, when <span class="ph">Nsight Systems</span>
                           detects that reports come from same target and that the same TSC value
                           corresponds to very close UTC times.
                           Targets are considered to be the same when either explicitly set environment variables
                           <samp class="ph codeph">NSYS_HW_ID</samp> are the same for both reports or when target hostnames are
                           the same and <samp class="ph codeph">NSYS_HW_ID</samp> is not set for either target.
                           The difference between UTC and TSC time offsets must be below 1 second to
                           choose TSC-based time synchronization.
                        </p>
                        <p class="p">
                           To find out which synchronization method was used, navigate to the <em class="ph i">Analysis Summary</em>
                           tab of an added report and check the <samp class="ph codeph">Report alignment source</samp> property
                           of a target. Note, that the first report won’t have this parameter.
                        </p>
                        <p class="p"><img class="image" src="graphics/report-alignment-source-TSC.png" alt="TODO"></img></p>
                        <p class="p"><img class="image" src="graphics/report-alignment-source-UTC.png" alt="TODO"></img></p>
                        <p class="p">When loading multiple reports into a single timeline, it is always advisable to first
                           check that time synchronization looks correct, by zooming into synchronization or
                           communication events that are expected to be aligned.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="multi-report-hierarchy"><a name="multi-report-hierarchy" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#multi-report-hierarchy" name="multi-report-hierarchy" shape="rect">23.2.&nbsp;Timeline Hierarchy</a></h3>
                     <div class="body">
                        <p class="p">
                           When reports are added to the same timeline <span class="ph">Nsight Systems</span>
                           will automatically line them up by timestamps as described above.
                           If you want <span class="ph">Nsight Systems</span> to also recognize matching
                           process or hardware information, you will need to set environment variables
                           <samp class="ph codeph">NSYS_SYSTEM_ID</samp> and <samp class="ph codeph">NSYS_HW_ID</samp> as shown
                           below at the time of report collection (such as when using
                           "<span class="ph">nsys</span> profile ..." command).
                        </p>
                        <p class="p">
                           When loading a pair of given report files into the same timeline,
                           they will be merged in one of the following configurations:
                        </p>
                        <ul class="ul">
                           <li class="li">
                              <p class="p">Different hardware — is used when reports are coming from
                                 different physical machines, and no hardware resources are shared in these reports.
                                 This mode is used when neither <samp class="ph codeph">NSYS_HW_ID</samp> or
                                 <samp class="ph codeph">NSYS_SYSTEM_ID</samp> is set and target hostnames are different
                                 or absent, and can be additionally signalled by specifying
                                 different <samp class="ph codeph">NSYS_HW_ID</samp> values.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">Different systems, same hardware — is used when reports are collected on different
                                 virtual machines (VMs) or containers on the same physical machine. To activate this mode,
                                 specify the same value of <samp class="ph codeph">NSYS_HW_ID</samp> when collecting the reports.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">Same system — is used when reports are collected within the same operating system
                                 (or container) environment. In this mode a process identifier (PID) 100 will refer to
                                 the same process in both reports.
                                 To manually activate this mode, specify the same value of
                                 <samp class="ph codeph">NSYS_SYSTEM_ID</samp> when collecting the reports.
                                 This mode is automatically selected when target hostnames are the same
                                 and neither <samp class="ph codeph">NSYS_HW_ID</samp> or <samp class="ph codeph">NSYS_SYSTEM_ID</samp>
                                 is provided.
                              </p>
                           </li>
                        </ul>
                        <p class="p">The following diagrams demonstrate typical cases:</p>
                        <p class="p"><img class="image" src="graphics/report-tiling-diagram.png" alt="TODO"></img></p>
                     </div>
                  </div>
                  <div class="topic nested1" id="multi-report-mpi-example"><a name="multi-report-mpi-example" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#multi-report-mpi-example" name="multi-report-mpi-example" shape="rect">23.3.&nbsp;Example: MPI</a></h3>
                     <div class="body">
                        <p class="p">A typical scenario is when a computing job is run using one of the MPI implementations.
                           Each instance of the app can be profiled separately, resulting in multiple report files.
                           For example:
                        </p><pre xml:space="preserve">
# Run MPI job without the profiler:
mpirun &lt;mpirun-options&gt; ./myApp
# Run MPI job and profile each instance of the application:
mpirun &lt;mpirun-options&gt; <span class="ph">nsys</span> profile -o report-%p &lt;<span class="ph">nsys</span>-options&gt;./myApp</pre><p class="p">When each MPI rank runs on a different node, the command above works fine,
                           since the default pairing mode (different hardware) will be used.
                        </p>
                        <p class="p">When all MPI ranks run the localhost only, use this command (value "A" was
                           chosen arbitrarily, it can be any non-empty string):
                        </p>
                        <p class="p"><samp class="ph codeph">NSYS_SYSTEM_ID=A mpirun &lt;mpirun-options&gt; <span class="ph">nsys</span> profile -o report-%p
                              &lt;<span class="ph">nsys</span>-options&gt; ./myApp</samp></p>
                        <p class="p">For convenience, the MPI rank can be encoded into the report filename.
                           For Open MPI, use the following command to create report files based on
                           the global rank value:
                        </p>
                        <p class="p"><samp class="ph codeph">mpirun &lt;mpirun-options&gt; <span class="ph">nsys</span> profile -o
                              report-%q{OMPI_COMM_WORLD_RANK} &lt;<span class="ph">nsys</span>-options&gt; ./myApp</samp></p>
                        <p class="p">MPICH-based implementations set the environment variable
                           <samp class="ph codeph">PMI_RANK</samp> and Slurm (<samp class="ph codeph">srun</samp>) provides the global MPI rank
                           in <samp class="ph codeph">SLURM_PROCID</samp>.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="multi-report-limitations"><a name="multi-report-limitations" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#multi-report-limitations" name="multi-report-limitations" shape="rect">23.4.&nbsp;Limitations</a></h3>
                     <div class="body">
                        <ul class="ul">
                           <li class="li">
                              <p class="p">Only report files collected with <span class="ph">Nsight Systems</span> version 2021.3
                                 and newer are fully supported.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">Sequential reports collected in a single CLI profiling session cannot
                                 be loaded into a single timeline yet.
                              </p>
                           </li>
                        </ul>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="expert-systems"><a name="expert-systems" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#expert-systems" name="expert-systems" shape="rect">24.&nbsp;Using Nsight Systems Expert System</a></h2>
                  <div class="body">
                     <p class="p">The <span class="ph">Nsight Systems</span> expert system is a feature
                        aimed at automatic detection of performance optimization
                        opportunities in an application's profile. It uses a set of predefined
                        rules to determine if the application has known bad patterns.
                     </p>
                  </div>
                  <div class="topic nested1" id="expert-systems-cli"><a name="expert-systems-cli" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#expert-systems-cli" name="expert-systems-cli" shape="rect">Using Expert System from the CLI</a></h3>
                     <div class="body">
                        <p class="p">usage:</p><pre xml:space="preserve"><span class="ph">nsys</span> [global-options] analyze [options]
       [nsys-rep-or-sqlite-file]</pre><p class="p">If a .nsys-rep file is given as the input file and there is no .sqlite
                           file with the same name in the same directory, it will be generated.
                        </p>
                        <p class="p"><strong class="ph b">Note:</strong> The Expert System view in the GUI will give you the
                           equivalent command line.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="expert-systems-gui"><a name="expert-systems-gui" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#expert-systems-gui" name="expert-systems-gui" shape="rect">Using Expert System from the GUI</a></h3>
                     <div class="body">
                        <p class="p">The Expert System View can be found in the same drop-down as the
                           Events View. If there is no .sqlite file with the same name as the .nsys-rep
                           file in the same directory, it will be generated.
                        </p>
                        <div class="p">The Expert System View has the following components:
                           
                           <ol class="ol">
                              <li class="li">Drop-down to select the rule to be run</li>
                              <li class="li">Rule description and advice summary</li>
                              <li class="li">CLI command that will give the same result</li>
                              <li class="li">Table containing results of running the rule</li>
                              <li class="li">Settings button that allows users to specify the rule’s arguments</li>
                           </ol>
                        </div>
                        <p class="p"><img class="image" src="graphics/expert-systems-gui.png" alt="Expert systems information as shown in the GUI"></img></p>
                        <div class="p">A context menu is available to correlate the table entry with the
                           timeline. The options are the same as the Events View:
                           
                           <ul class="ul">
                              <li class="li">Zoom to Selected on Timeline (ctrl+double-click)</li>
                           </ul>
                        </div>
                        <p class="p">The highlighting is not supported for rules that do not return an
                           event but rather an arbitrary time range (e.g. GPU utilization rules).
                        </p>
                        <p class="p">The CLI and GUI share the same rule scripts and messages. There might
                           be some formatting differences between the output table in GUI and CLI.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="expert-systems-rules"><a name="expert-systems-rules" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#expert-systems-rules" name="expert-systems-rules" shape="rect">Expert System Rules</a></h3>
                     <div class="body">
                        <p class="p">Rules are scripts that run on the SQLite DB output
                           from <span class="ph">Nsight Systems</span> to find common improvable usage patterns.
                           
                        </p>
                        <p class="p">Each rule has an advice summary with explanation of the problem found
                           and suggestions to address it. Only the top 50 results are displayed by
                           default.
                        </p>
                        <p class="p">There are currently six rules in the expert system. They are
                           described below. Additional rules will be made available in a future 
                           version of <span class="ph">Nsight Systems</span>. 
                        </p>
                     </div>
                     <div class="topic nested2" id="expert-systems-synchronous-operations"><a name="expert-systems-synchronous-operations" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#expert-systems-synchronous-operations" name="expert-systems-synchronous-operations" shape="rect">CUDA Synchronous Operation Rules</a></h3>
                        <div class="body">
                           <p class="p"><strong class="ph b">Asynchronous memcpy with pageable memory</strong></p>
                           <p class="p">This rule identifies asynchronous memory transfers that end up
                              becoming synchronous if the memory is pageable. This rule is not
                              applicable for <span class="ph">Nsight Systems Embedded Platforms Edition</span></p>
                           <p class="p">Suggestion: If applicable, use pinned memory instead</p>
                           <p class="p"><img class="image" src="graphics/expert-systems-pageable.png" alt="Asynchronous memcpy with pageable memory"></img></p>
                           <p class="p"><strong class="ph b">Synchronous Memcpy</strong></p>
                           <p class="p">This rule identifies synchronous memory transfers that block the
                              host.
                           </p>
                           <p class="p">Suggestion: Use cudaMemcpy*Async APIs instead.</p>
                           <p class="p"><strong class="ph b">Synchronous Memset</strong></p>
                           <p class="p">This rule identifies synchronous memset operations that block the
                              host.
                           </p>
                           <p class="p">Suggestion: Use cudaMemset*Async APIs instead.</p>
                           <p class="p"><strong class="ph b">Synchronization APIs</strong></p>
                           <p class="p">This rule identifies synchronization APIs that block the host until
                              all issued CUDA calls are complete.
                           </p>
                           <p class="p">Suggestions: Avoid excessive use of synchronization. Use
                              asynchronous CUDA event calls, such as cudaStreamWaitEvent and
                              cudaEventSynchronize, to prevent host synchronization.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="expert-systems-GPU-utilization"><a name="expert-systems-GPU-utilization" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#expert-systems-GPU-utilization" name="expert-systems-GPU-utilization" shape="rect">GPU Low Utilization Rules</a></h3>
                        <div class="body">
                           <p class="p"><span class="ph">Nsight Systems</span> determines GPU utilization based on API
                              trace data in the collection. Current rules consider CUDA, Vulkan,
                              DX12, and OpenGL API use of the GPU.
                           </p>
                           <p class="p"><strong class="ph b">GPU Starvation</strong></p>
                           <p class="p">This rule identifies time ranges where a GPU is idle for longer
                              than 500ms. The threshold is adjustable.
                           </p>
                           <p class="p">Suggestions: Use CPU sampling data, OS Runtime blocked state
                              backtraces, and/or OS Runtime APIs related to thread synchronization
                              to understand if a sluggish or blocked CPU is causing the gaps. Add
                              NVTX annotations to CPU code to understand the reason behind the gaps.
                           </p>
                           <p class="p">Notes:</p>
                           <ul class="ul">
                              <li class="li">For each process, each GPU is examined, and gaps are found within
                                 the time range that starts with the beginning of the first GPU
                                 operation on that device and ends with the end of the last GPU
                                 operation on that device.
                              </li>
                              <li class="li">GPU gaps that cannot be addressed by the user are excluded.
                                 This includes:
                                 
                                 <ul class="ul">
                                    <li class="li">Profiling overhead in the middle of a GPU gap.</li>
                                    <li class="li">The initial gap in the report that is seen before the first GPU
                                       operation.
                                    </li>
                                    <li class="li">The final gap that is seen after the last GPU operation.</li>
                                 </ul>
                              </li>
                           </ul>
                           <p class="p"><strong class="ph b">GPU Low Utilization</strong></p>
                           <p class="p">This rule identifies time regions with low utilization.</p>
                           <p class="p">Suggestions: Use CPU sampling data, OS Runtime blocked state
                              backtraces, and/or OS Runtime APIs related to thread synchronization
                              to understand if a sluggish or blocked CPU is causing the gaps. Add
                              NVTX annotations to CPU code to understand the reason behind the gaps.
                           </p>
                           <p class="p">Notes:</p>
                           <ul class="ul">
                              <li class="li">For each process, each GPU is examined, and gaps are found within
                                 the time range that starts with the beginning of the first GPU
                                 operation on that device and ends with the end of the last GPU
                                 operation on that device. This time range is then divided into equal
                                 chunks, and the GPU utilization is calculated for each chunk. The
                                 utilization includes all GPU operations as well as profiling
                                 overheads that the user cannot address.
                              </li>
                              <li class="li">The utilization refers to the "time" utilization and not the
                                 "resource" utilization. This rule attempts to find time gaps when the
                                 GPU is or isn't being used, but does not take into account how many
                                 GPU resources are being used. Therefore, a single running memcpy is
                                 considered the same amount of "utilization" as a huge kernel that
                                 takes over all the cores. If multiple operations run concurrently in
                                 the same chunk, their utilization will be added up and may exceed
                                 100%.
                              </li>
                              <li class="li">Chunks with an in-use percentage less than the threshold value are
                                 displayed. If consecutive chunks have a low in-use percentage, the
                                 individual chunks are coalesced into a single display record, keeping
                                 the weighted average of percentages. This is why returned chunks may
                                 have different durations.
                              </li>
                           </ul>
                        </div>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="import-nvtxt"><a name="import-nvtxt" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#import-nvtxt" name="import-nvtxt" shape="rect">25.&nbsp;Import NVTXT</a></h2>
                  <div class="body">
                     <p class="p"><strong class="ph b">ImportNvtxt</strong> is an utility which allows conversion of a <a class="xref" href="https://docs.nvidia.com/gameworks/index.html#gameworkslibrary/nvtx/analysis_nvtxt_file_extension.htm" target="_blank" shape="rect">NVTXT</a> file to a
                        <span class="ph">Nsight Systems</span> report file (*.nsys-rep) or to merge it with an
                        existing report file.
                     </p>
                     <p class="p"><strong class="ph b">Note</strong>: NvtxtImport supports custom <strong class="ph b">TimeBase</strong> values. Only these
                        values are supported:
                     </p>
                     <ul class="ul">
                        <li class="li">
                           <p class="p"><strong class="ph b">Manual</strong> — timestamps are set using absolute values.
                           </p>
                        </li>
                        <li class="li">
                           <p class="p"><strong class="ph b">Relative</strong> — timestamps are set using relative values with regards
                              to report file which is being merged with nvtxt file.
                           </p>
                        </li>
                        <li class="li">
                           <p class="p"><strong class="ph b">ClockMonotonicRaw</strong> — timestamps values in nvtxt file are considered
                              to be gathered on the same target as the report file which is to be merged
                              with nvtxt using <samp class="ph codeph">clock_gettime(CLOCK_MONOTONIC_RAW, ...)</samp>
                              call.
                           </p>
                        </li>
                        <li class="li">
                           <p class="p"><strong class="ph b">CNTVCT</strong> — timestamps values in nvtxt file are considered to be
                              gathered on the same target as the report file which is to be merged with
                              nvtxt using CNTVCT values.
                           </p>
                        </li>
                     </ul>
                     <p class="p">You can get usage info via help message:</p>
                     <p class="p">Print help message:</p><pre xml:space="preserve">-h [ --help ]</pre><p class="p">Show information about report file:</p><pre xml:space="preserve">--cmd info -i [--input] arg</pre><p class="p">Create report file from existing nvtxt file:</p><pre xml:space="preserve">--cmd create -n [--nvtxt] arg -o [--output] arg [-m [--mode] mode_name mode_args] [--target &lt;Hw:Vm&gt;] [--update_report_time]</pre><p class="p">Merge nvtxt file to existing report file:</p><pre xml:space="preserve">--cmd merge -i [--input] arg -n [--nvtxt] arg -o [--output] arg [-m [--mode] mode_name mode_args] [--target &lt;Hw:Vm&gt;] [--update_report_time]</pre><div class="p">Modes description:
                        
                        <ul class="ul">
                           <li class="li">
                              <p class="p">lerp - Insert with linear interpolation</p><pre xml:space="preserve">--mode lerp --ns_a arg --ns_b arg [--nvtxt_a arg --nvtxt_b arg]</pre></li>
                           <li class="li">
                              <p class="p">lin - insert with linear equation</p><pre xml:space="preserve">--mode lin  --ns_a arg --freq arg [--nvtxt_a arg]</pre></li>
                        </ul>
                     </div>
                     <div class="p">Modes' parameters:
                        
                        <ul class="ul">
                           <li class="li">
                              <p class="p">ns_a - a nanoseconds value</p>
                           </li>
                           <li class="li">
                              <p class="p">ns_b - a nanoseconds value (greater than ns_a)</p>
                           </li>
                           <li class="li">
                              <p class="p">nvtxt_a - an nvtxt file's time unit value corresponding to ns_a nanoseconds</p>
                           </li>
                           <li class="li">
                              <p class="p">nvtxt_b - an nvtxt file's time unit value corresponding to ns_b nanoseconds</p>
                           </li>
                           <li class="li">
                              <p class="p">freq    - the nvtxt file's timer frequency</p>
                           </li>
                           <li class="li">
                              <p class="p">--target &lt;Hw:Vm&gt; - specify target id, e.g. --target 0:1</p>
                           </li>
                           <li class="li">
                              <p class="p">--update_report_time - prolong report's profiling session time while
                                 merging if needed. Without this option all events outside the
                                 profiling session time window will be skipped during merging.
                              </p>
                           </li>
                        </ul>
                     </div>
                  </div>
                  <div class="topic nested1" id="nvtxt-commands"><a name="nvtxt-commands" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#nvtxt-commands" name="nvtxt-commands" shape="rect">Commands</a></h3>
                     <div class="body">
                        <p class="p"><strong class="ph b">Info</strong></p>
                        <p class="p">To find out report's start and end time use <strong class="ph b">info</strong> command.
                        </p>
                        <p class="p">Usage:</p><pre xml:space="preserve">ImportNvtxt --cmd info -i [--input] arg</pre><p class="p">Example:</p><pre xml:space="preserve">ImportNvtxt info Report.nsys-rep
Analysis start (ns)	83501026500000
Analysis end (ns)	83506375000000</pre><p class="p"><strong class="ph b">Create</strong></p>
                        <p class="p">You can create a report file using existing NVTXT with <strong class="ph b">create</strong>
                           command.
                        </p>
                        <p class="p">Usage:</p><pre xml:space="preserve">ImportNvtxt --cmd create -n [--nvtxt] arg -o [--output] arg [-m [--mode] mode_name mode_args]</pre><div class="p">Available modes are:
                           
                           <ul class="ul">
                              <li class="li">
                                 <p class="p"><strong class="ph b">lerp</strong> — insert with linear interpolation.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><strong class="ph b">lin</strong> — insert with linear equation.
                                 </p>
                              </li>
                           </ul>
                        </div>
                        <p class="p">Usage for <strong class="ph b">lerp</strong> mode is:
                        </p><pre xml:space="preserve">--mode lerp --ns_a arg --ns_b arg [--nvtxt_a arg --nvtxt_b arg]</pre><div class="p">with:
                           
                           <ul class="ul">
                              <li class="li">
                                 <p class="p"><samp class="ph codeph">ns_a</samp> — a nanoseconds value.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><samp class="ph codeph">ns_b</samp> — a nanoseconds value (greater than <samp class="ph codeph">ns_a</samp>).
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><samp class="ph codeph">nvtxt_a</samp> — an nvtxt file's time unit value
                                    corresponding to <samp class="ph codeph">ns_a</samp> nanoseconds.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><samp class="ph codeph">nvtxt_b</samp> — an nvtxt file's time unit value
                                    corresponding to <samp class="ph codeph">ns_b</samp> nanoseconds.
                                 </p>
                              </li>
                           </ul>
                        </div>
                        <p class="p">If <samp class="ph codeph">nvtxt_a</samp> and <samp class="ph codeph">nvtxt_b</samp> are not
                           specified, they are respectively set to nvtxt file's minimum and maximum
                           time value.
                        </p>
                        <p class="p">Usage for <strong class="ph b">lin</strong> mode is:
                        </p><pre xml:space="preserve">--mode lin --ns_a arg --freq arg [--nvtxt_a arg]</pre><div class="p">with:
                           
                           <ul class="ul">
                              <li class="li">
                                 <p class="p"><samp class="ph codeph">ns_a</samp> — a nanoseconds value.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><samp class="ph codeph">freq</samp> — the nvtxt file's timer frequency.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><samp class="ph codeph">nvtxt_a</samp> — an nvtxt file's time unit value
                                    corresponding to <samp class="ph codeph">ns_a</samp> nanoseconds.
                                 </p>
                              </li>
                           </ul>
                        </div>
                        <p class="p">If <samp class="ph codeph">nvtxt_a</samp> is not specified, it is set to nvtxt
                           file's minimum time value.
                        </p>
                        <p class="p"><strong class="ph b">Examples:</strong></p><pre xml:space="preserve">ImportNvtxt --cmd create -n Sample.nvtxt -o Report.nsys-rep</pre><p class="p">
                           The output will be a new generated report file which can be opened and
                           viewed by <span class="ph">Nsight Systems</span>.
                        </p>
                        <p class="p"><strong class="ph b">Merge</strong></p>
                        <p class="p">To merge NVTXT file with an existing report file use <strong class="ph b">merge</strong>
                           command.
                        </p>
                        <p class="p">Usage:</p><pre xml:space="preserve">ImportNvtxt --cmd merge -i [--input] arg -n [--nvtxt] arg -o [--output] arg [-m [--mode] mode_name mode_args]</pre><div class="p">Available modes are:
                           
                           <ul class="ul">
                              <li class="li">
                                 <p class="p"><strong class="ph b">lerp</strong> — insert with linear interpolation.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><strong class="ph b">lin</strong> — insert with linear equation.
                                 </p>
                              </li>
                           </ul>
                        </div>
                        <p class="p">Usage for <strong class="ph b">lerp</strong> mode is:
                        </p><pre xml:space="preserve">--mode lerp --ns_a arg --ns_b arg [--nvtxt_a arg --nvtxt_b arg]</pre><div class="p">with:
                           
                           <ul class="ul">
                              <li class="li">
                                 <p class="p"><samp class="ph codeph">ns_a</samp> — a nanoseconds value.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><samp class="ph codeph">ns_b</samp> — a nanoseconds value (greater than <samp class="ph codeph">ns_a</samp>).
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><samp class="ph codeph">nvtxt_a</samp> — an nvtxt file's time unit value
                                    corresponding to <samp class="ph codeph">ns_a</samp> nanoseconds.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><samp class="ph codeph">nvtxt_b</samp> — an nvtxt file's time unit value
                                    corresponding to <samp class="ph codeph">ns_b</samp> nanoseconds.
                                 </p>
                              </li>
                           </ul>
                        </div>
                        <p class="p">If <samp class="ph codeph">nvtxt_a</samp> and <samp class="ph codeph">nvtxt_b</samp> are not
                           specified, they are respectively set to nvtxt file's minimum and maximum
                           time value.
                        </p>
                        <p class="p">Usage for <strong class="ph b">lin</strong> mode is:
                        </p><pre xml:space="preserve">--mode lin  --ns_a arg --freq arg [--nvtxt_a arg]</pre><div class="p">with:
                           
                           <ul class="ul">
                              <li class="li">
                                 <p class="p"><samp class="ph codeph">ns_a</samp> — a nanoseconds value.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><samp class="ph codeph">freq</samp> — the nvtxt file's timer frequency.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p"><samp class="ph codeph">nvtxt_a</samp> — an nvtxt file's time unit value
                                    corresponding to <samp class="ph codeph">ns_a</samp> nanoseconds.
                                 </p>
                              </li>
                           </ul>
                        </div>
                        <p class="p">If <samp class="ph codeph">nvtxt_a</samp> is not specified, it is set to nvtxt
                           file's minimum time value.
                        </p>
                        <p class="p">Time values in <samp class="ph codeph">&lt;filename.nvtxt&gt;</samp> are assumed to
                           be nanoseconds if no mode specified.
                        </p>
                        <p class="p">Example</p><pre xml:space="preserve">ImportNvtxt --cmd merge -i Report.nsys-rep -n Sample.nvtxt -o NewReport.nsys-rep</pre><p class="p"></p>
                        <p class="p hr"></p>
                        <p class="p"></p>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="ballista-overview"><a name="ballista-overview" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#ballista-overview" name="ballista-overview" shape="rect">Visual Studio Integration</a></h2>
                  <div class="body">
                     <p class="p"><span class="ph">NVIDIA Nsight Integration</span> is a <span class="ph">Visual Studio</span> extension that allows you to
                        access the power of <span class="ph">Nsight Systems</span> from within <span class="ph">Visual Studio</span>.
                     </p>
                     <p class="p">
                        When <span class="ph">Nsight Systems</span> is installed along with <span class="ph">NVIDIA Nsight Integration</span>,
                        <span class="ph">Nsight Systems</span> activities will appear under 
                        the NVIDIA Nsight menu in the <span class="ph">Visual Studio</span> menu bar. These activities launch
                        <span class="ph">Nsight Systems</span> with the current project settings and executable.
                     </p>
                     <p class="p"><img class="image" src="graphics/visual_studio_extension_menu_trace_command.png" alt="Install extension to Microsoft Visual Studio"></img></p>
                     <div class="p">
                        Selecting the "Trace" command will launch <span class="ph">Nsight Systems</span>,
                        create a new <span class="ph">Nsight Systems</span> project and apply settings from
                        the current <span class="ph">Visual Studio</span> project:
                        
                        <ul class="ul">
                           <li class="li">
                              <p class="p">Target application path</p>
                           </li>
                           <li class="li">
                              <p class="p">Command line parameters</p>
                           </li>
                           <li class="li">
                              <p class="p">Working folder</p>
                           </li>
                        </ul>
                     </div>
                     <p class="p">
                        If the "Trace" command has already been used with this <span class="ph">Visual Studio</span> project
                        then <span class="ph">Nsight Systems</span> will load the respective
                        <span class="ph">Nsight Systems</span> project and any previously captured trace
                        sessions will be available for review using the <span class="ph">Nsight Systems</span>
                        project explorer tree.
                     </p>
                     <div class="p">
                        For more information about using <span class="ph">Nsight Systems</span> from 
                        within <span class="ph">Visual Studio</span>, please visit
                        
                        <ul class="ul">
                           <li class="li">
                              <p class="p"><a class="xref" href="https:/developer.nvidia.com/nsight-tools-visual-studio-integration" target="_blank" shape="rect">NVIDIA Nsight Integration Overview</a></p>
                           </li>
                           <li class="li">
                              <p class="p"><a class="xref" href="https:/docs.nvidia.com/nsight-vs-integration/index.html" target="_blank" shape="rect">NVIDIA Nsight Integration User Guide</a></p>
                           </li>
                        </ul>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="troubleshooting"><a name="troubleshooting" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#troubleshooting" name="troubleshooting" shape="rect">27.&nbsp;Troubleshooting</a></h2>
                  <div class="body"></div>
                  <div class="topic nested1" id="general-troubleshooting"><a name="general-troubleshooting" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#general-troubleshooting" name="general-troubleshooting" shape="rect">27.1.&nbsp;General Troubleshooting</a></h3>
                     <div class="body">
                        <p class="p"><strong class="ph b">Profiling</strong></p>
                        <div class="p">
                           If the profiler behaves unexpectedly during the profiling session, or the
                           profiling session fails to start, try the following steps:
                           
                           <ul class="ul">
                              <li class="li">Close the host application.</li>
                              <li class="li">Restart the target device.</li>
                              <li class="li">Start the host application and connect to the target device.</li>
                           </ul>
                        </div>
                        <p class="p"><span class="ph">Nsight Systems</span> uses a settings file
                           (<samp class="ph codeph">NVIDIA Nsight Systems.ini</samp>) on the host to store
                           information about loaded projects, report files, window layout
                           configuration, etc.  Location of the settings file is described in the
                           <strong class="ph b">Help → About</strong> dialog.  Deleting the settings file will restore
                           <span class="ph">Nsight Systems</span> to a fresh state, but all projects and
                           reports will disappear from the Project Explorer.
                        </p>
                        <p class="p"><strong class="ph b">Environment Variables</strong></p>
                        <div class="p">By default, <span class="ph">Nsight Systems</span> writes temporary files to
                           <samp class="ph codeph">/tmp</samp> directory. If you are using a system that does not
                           allow writing to <samp class="ph codeph">/tmp</samp> or where the <samp class="ph codeph">/tmp</samp>
                           directory has limited storage you can use the TMPDIR
                           environment variable to set a different location. An example:
                           <pre xml:space="preserve">TMPDIR=/testdata ./bin/nsys profile -t cuda matrixMul
       </pre></div>
                        <p class="p">Environment variable control support for Windows target trace is not
                           available, but there is a quick workaround:
                        </p>
                        <ul class="ul">
                           <li class="li">Create a batch file that sets the env vars and launches your
                              application.
                           </li>
                           <li class="li">Set <span class="ph">Nsight Systems</span> to launch the batch file as its
                              target, i.e. set the project settings target path to the path of batch
                              file.
                           </li>
                           <li class="li">Start the trace. <span class="ph">Nsight Systems</span> will launch the batch
                              file in a new cmd instance and trace any child process it launches. In
                              fact, it will trace the whole process tree whose root is the cmd running
                              your batch file.
                           </li>
                        </ul>
                        <p class="p"><strong class="ph b">WebGL Testing</strong></p>
                        <div class="p"><span class="ph">Nsight Systems</span> cannot profile using the default Chrome
                           launch command. To profile WebGL please follow the following command
                           structure:
                           <pre xml:space="preserve">“C:\Program Files (x86)\Google\Chrome\Application\chrome.exe”
       --inprocess-gpu --no-sandbox --disable-gpu-watchdog --use-angle=gl
       https://webglsamples.org/aquarium/aquarium.html
       </pre></div>
                        <p class="p"><strong class="ph b">Common Issues with QNX Targets</strong></p>
                        <ul class="ul">
                           <li class="li">
                              <p class="p">Make sure that <samp class="ph codeph">tracelogger</samp> utility is available and
                                 can be run on the target.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">Make sure that <samp class="ph codeph">/tmp</samp> directory is accessible and supports
                                 sub-directories.
                              </p>
                           </li>
                           <li class="li">
                              <p class="p">When switching between <span class="ph">Nsight Systems</span> versions, processes
                                 related to the previous version, including profiled applications forked
                                 by the daemon, must be killed before the new version is used. If you
                                 experience issues after switching between <span class="ph">Nsight Systems</span>
                                 versions, try rebooting the target.
                              </p>
                           </li>
                        </ul>
                     </div>
                  </div>
                  <div class="topic nested1" id="cli-troubleshooting"><a name="cli-troubleshooting" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#cli-troubleshooting" name="cli-troubleshooting" shape="rect">CLI Troubleshooting</a></h3>
                     <div class="body">
                        <p class="p">
                           If you have collected a report file using the CLI and the report will not
                           open in the GUI, check to see that your GUI version is the same or
                           greater than the CLI version you used. If it is not, download a new
                           version of the <span class="ph">Nsight Systems</span> GUI and you will be able to
                           load and visualize your report.
                        </p>
                        <p class="p">
                           This situation occurs most frequently when you update
                           <span class="ph">Nsight Systems</span> using a CLI only package, such as the package
                           available from the NVIDIA HPC SDK.
                        </p>
                     </div>
                  </div>
                  <div class="topic nested1" id="launch-stopped-linux"><a name="launch-stopped-linux" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#launch-stopped-linux" name="launch-stopped-linux" shape="rect">27.3.&nbsp;Launch Processes in Stopped State</a></h3>
                     <div class="body">
                        <p class="p">
                           In many cases, it is important to profile an application from the very
                           beginning of its execution. When launching processes,
                           <span class="ph">Nsight Systems</span> takes care of it by making sure that the
                           profiling session is fully initialized before making the
                           <samp class="ph codeph">exec()</samp> system call on Linux.
                        </p>
                        <p class="p">
                           If the process launch capabilities of <span class="ph">Nsight Systems</span> are not
                           sufficient, the application should be launched manually, and the profiler
                           should be configured to attach to the already launched process. One approach
                           would be to call <samp class="ph codeph">sleep()</samp> somewhere early in the application
                           code, which would provide time for the user to attach to the process in
                           <span class="ph">Nsight Systems Embedded Platforms Edition</span>, but there are two other more convenient
                           mechanisms that can be used on Linux, without the need to recompile the
                           application. (Note that the rest of this section is only applicable to
                           Linux-based target devices.)
                        </p>
                        <p class="p">
                           Both mechanisms ensure that between the time the process is created (and
                           therefore its PID is known) and the time any of the application's code is
                           called, the process is stopped and waits for a signal to be delivered before
                           continuing.
                        </p>
                     </div>
                     <div class="topic nested2" id="ld-preload"><a name="ld-preload" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#ld-preload" name="ld-preload" shape="rect">LD_PRELOAD</a></h3>
                        <div class="body">
                           <p class="p">
                              The first mechanism uses <samp class="ph codeph">LD_PRELOAD</samp> environment
                              variable. It only works with dynamically linked binaries, since static
                              binaries do not invoke the runtime linker, and therefore are not affected
                              by the <samp class="ph codeph">LD_PRELOAD</samp> environment variable.
                           </p>
                           <ul class="ul">
                              <li class="li">
                                 <p class="p">For ARMv7 binaries, preload</p><pre xml:space="preserve">/opt/nvidia/nsight_systems/libLauncher32.so</pre></li>
                              <li class="li">
                                 <p class="p">Otherwise if running from host, preload</p><pre xml:space="preserve">/opt/nvidia/nsight_systems/libLauncher64.so</pre></li>
                              <li class="li">
                                 <p class="p">Otherwise if running from CLI, preload</p><pre xml:space="preserve">[installation_directory]/libLauncher64.so</pre></li>
                           </ul>
                           <p class="p">
                              The most common way to do that is to specify the environment variable as
                              part of the process launch command, for example:
                           </p><pre xml:space="preserve">$ LD_PRELOAD=/opt/nvidia/nsight_systems/libLauncher64.so ./my-aarch64-binary --arguments</pre><p class="p">
                              When loaded, this library will send itself a <samp class="ph codeph">SIGSTOP</samp>
                              signal, which is equivalent to typing <samp class="ph codeph">Ctrl+Z</samp> in the
                              terminal. The process is now a background job, and you can use standard
                              commands like jobs, <samp class="ph codeph">fg</samp> and <samp class="ph codeph">bg</samp> to
                              control them. Use <samp class="ph codeph">jobs -l</samp> to see the PID of the launched
                              process.
                           </p>
                           <p class="p">
                              When attaching to a stopped process, <span class="ph">Nsight Systems</span> will
                              send <samp class="ph codeph">SIGCONT</samp> signal, which is equivalent to using the
                              <samp class="ph codeph">bg</samp> command.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="launcher"><a name="launcher" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#launcher" name="launcher" shape="rect">Launcher</a></h3>
                        <div class="body">
                           <p class="p">
                              The second mechanism can be used with any binary. Use
                              <samp class="ph codeph">[installation_directory]/launcher</samp> to launch your
                              application, for example:
                           </p><pre xml:space="preserve">$ /opt/nvidia/nsight_systems/launcher ./my-binary --arguments</pre><p class="p">The process will be launched, daemonized, and wait for
                              <samp class="ph codeph">SIGUSR1</samp> signal. After attaching to the process with
                              <span class="ph">Nsight Systems</span>, the user needs to manually resume
                              execution of the process from command line:
                           </p><pre xml:space="preserve">$ pkill -USR1 launcher</pre><div class="p">
                              <div class="tablenoborder">
                                 <table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
                                    <tbody class="tbody">
                                       <tr class="row">
                                          <td class="entry nvidiagreen" align="center" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                             <p class="p">&nbsp;&nbsp;<strong class="ph b">Note:</strong>&nbsp;&nbsp;
                                             </p>
                                          </td>
                                          <td class="entry notebox" valign="middle" width="NaN%" rowspan="1" colspan="1">
                                             <p class="p">Note that <samp class="ph codeph">pkill</samp> will send the signal to any process with
                                                the matching name. If that is not desirable, use <samp class="ph codeph">kill</samp> to
                                                send it to a specific process. The standard output and error streams are
                                                redirected to <samp class="ph codeph">/tmp/stdout_&lt;PID&gt;.txt</samp> and
                                                <samp class="ph codeph">/tmp/stderr_&lt;PID&gt;.txt</samp>.
                                             </p>
                                          </td>
                                       </tr>
                                    </tbody>
                                 </table>
                              </div>
                           </div>
                           <p class="p">The launcher mechanism is more complex and less automated than the
                              LD_PRELOAD option, but gives more control to the user.
                           </p>
                        </div>
                     </div>
                  </div>
                  <div class="topic nested1" id="gui-troubleshooting"><a name="gui-troubleshooting" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#gui-troubleshooting" name="gui-troubleshooting" shape="rect">GUI Troubleshooting</a></h3>
                     <div class="body">
                        <p class="p">
                           If opening the <span class="ph">Nsight Systems</span> Linux GUI fails with one of the
                           following errors, you may be missing some required libraries:
                        </p><pre xml:space="preserve">This application failed to start because it could not find or load the Qt platform plugin "xcb" in "". Available platform plugins are: xcb. Reinstalling the application may fix this problem.</pre><p class="p">
                           or
                        </p><pre xml:space="preserve">error while loading shared libraries: [library_name]: cannot open shared object file: No such file or directory</pre></div>
                     <div class="topic nested2" id="gui-troubleshooting-root-ubuntu-centos"><a name="gui-troubleshooting-root-ubuntu-centos" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#gui-troubleshooting-root-ubuntu-centos" name="gui-troubleshooting-root-ubuntu-centos" shape="rect">Ubuntu 18.04/20.04 and CentOS 7/8 with root privileges</a></h3>
                        <div class="body">
                           <div class="p">
                              <ul class="ul">
                                 <li class="li">Launch the following command, which will install all the required libraries in system directories:
                                    <pre xml:space="preserve">[installation_path]/host-linux-x64/Scripts/install-dependencies.sh</pre></li>
                                 <li class="li">Launch the Linux GUI as usual.</li>
                              </ul>
                           </div>
                        </div>
                     </div>
                     <div class="topic nested2" id="gui-troubleshooting-non-root-ubuntu-centos"><a name="gui-troubleshooting-non-root-ubuntu-centos" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#gui-troubleshooting-non-root-ubuntu-centos" name="gui-troubleshooting-non-root-ubuntu-centos" shape="rect">Ubuntu 18.04/20.04 and CentOS 7/8 without root privileges</a></h3>
                        <div class="body">
                           <div class="p">
                              <ul class="ul">
                                 <li class="li">Choose the directory where dependencies will be installed (<samp class="ph codeph">dependencies_path</samp>). This directory should be writeable for the current user.
                                 </li>
                                 <li class="li">Launch the following command (if it has already been run, move to the next step), which will install all the required libraries
                                    in <samp class="ph codeph">[dependencies_path]</samp>:
                                    <pre xml:space="preserve">[installation_path]/host-linux-x64/Scripts/install-dependencies-without-root.sh [dependencies_path]</pre></li>
                                 <li class="li">Further, use the following command to launch the Linux GUI:
                                    <pre xml:space="preserve">source [installation_path]/host-linux-x64/Scripts/setup-dependencies-environment.sh [dependencies_path] &amp;&amp; [installation_path]/host-linux-x64/nsys-ui</pre></li>
                              </ul>
                           </div>
                        </div>
                     </div>
                     <div class="topic nested2" id="gui-troubleshooting-other"><a name="gui-troubleshooting-other" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#gui-troubleshooting-other" name="gui-troubleshooting-other" shape="rect">Other platforms, or if the previous steps did not help</a></h3>
                        <div class="body">
                           <p class="p">
                              Launch <span class="ph">Nsight Systems</span> using the following command line to
                              determine which libraries are missing and install them.
                           </p><pre xml:space="preserve">$ QT_DEBUG_PLUGINS=1 ./nsys-ui</pre><p class="p">
                              If the workload does not run when launched via
                              <span class="ph">Nsight Systems</span> or the timeline is empty, check the
                              stderr.log and stdout.log (click on drop-down menu showing
                              <strong class="ph b">Timeline View</strong> and click on <strong class="ph b">Files</strong>) to see the errors
                              encountered by the app.
                           </p>
                           <blockquote class="lq">
                              <p class="p"><img class="image" src="graphics/stderr-log.png" alt="Stderr Log"></img></p>
                           </blockquote>
                        </div>
                     </div>
                  </div>
                  <div class="topic nested1" id="symbol-resolution-troubleshooting"><a name="symbol-resolution-troubleshooting" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#symbol-resolution-troubleshooting" name="symbol-resolution-troubleshooting" shape="rect">Symbol Resolution</a></h3>
                     <div class="body">
                        <p class="p">
                           If stack trace information is missing symbols and you have a symbol file,
                           you can manually re-resolve using the ResolveSymbols utility. This can be
                           done by right-clicking the report file in the Project Explorer window and
                           selecting "Resolve Symbols...".
                           
                        </p>
                        <p class="p">Alternatively, you can find the utility as a separate executable in the
                           <samp class="ph codeph">[installation_path]\Host</samp> directory.
                           This utility works with ELF format files, with Windows PDB directories
                           and symbol servers, or with files where each line is in the format
                           <samp class="ph codeph">&lt;start&gt;&lt;length&gt;&lt;name&gt;</samp>.
                           
                        </p>
                        <table cellpadding="4" cellspacing="0" summary="" border="1" class="simpletable">
                           <tr class="sthead">
                              <th valign="bottom" align="left" id="d54e16503" class="stentry" width="10%" rowspan="1" colspan="1">Short</th>
                              <th valign="bottom" align="left" id="d54e16506" class="stentry" width="20%" rowspan="1" colspan="1">Long</th>
                              <th valign="bottom" align="left" id="d54e16509" class="stentry" width="20%" rowspan="1" colspan="1">Argument</th>
                              <th valign="bottom" align="left" id="d54e16512" class="stentry" width="50%" rowspan="1" colspan="1">Description</th>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e16503" class="stentry" rowspan="1" colspan="1">-h</td>
                              <td valign="top" headers="d54e16506" class="stentry" rowspan="1" colspan="1">--help</td>
                              <td valign="top" headers="d54e16509" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                              <td valign="top" headers="d54e16512" class="stentry" rowspan="1" colspan="1">Help message providing information about available options.</td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e16503" class="stentry" rowspan="1" colspan="1">-l</td>
                              <td valign="top" headers="d54e16506" class="stentry" rowspan="1" colspan="1">--process-list</td>
                              <td valign="top" headers="d54e16509" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                              <td valign="top" headers="d54e16512" class="stentry" rowspan="1" colspan="1">Print global process IDs list</td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e16503" class="stentry" rowspan="1" colspan="1">-s</td>
                              <td valign="top" headers="d54e16506" class="stentry" rowspan="1" colspan="1">--sym-file</td>
                              <td valign="top" headers="d54e16509" class="stentry" rowspan="1" colspan="1">filename</td>
                              <td valign="top" headers="d54e16512" class="stentry" rowspan="1" colspan="1">Path to symbol file</td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e16503" class="stentry" rowspan="1" colspan="1">-b</td>
                              <td valign="top" headers="d54e16506" class="stentry" rowspan="1" colspan="1">--base-addr</td>
                              <td valign="top" headers="d54e16509" class="stentry" rowspan="1" colspan="1">address</td>
                              <td valign="top" headers="d54e16512" class="stentry" rowspan="1" colspan="1">If set then &lt;start&gt; in symbol file is treated as
                                 relative address starting from this base address
                              </td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e16503" class="stentry" rowspan="1" colspan="1">-p</td>
                              <td valign="top" headers="d54e16506" class="stentry" rowspan="1" colspan="1">--global-pid</td>
                              <td valign="top" headers="d54e16509" class="stentry" rowspan="1" colspan="1">pid</td>
                              <td valign="top" headers="d54e16512" class="stentry" rowspan="1" colspan="1">Which process in the report should be resolved. May be
                                 omitted if there is only one process in the report.
                              </td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e16503" class="stentry" rowspan="1" colspan="1">-f</td>
                              <td valign="top" headers="d54e16506" class="stentry" rowspan="1" colspan="1">--force</td>
                              <td valign="top" headers="d54e16509" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                              <td valign="top" headers="d54e16512" class="stentry" rowspan="1" colspan="1">This option forces use of a given symbol file.</td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e16503" class="stentry" rowspan="1" colspan="1">-i</td>
                              <td valign="top" headers="d54e16506" class="stentry" rowspan="1" colspan="1">--report</td>
                              <td valign="top" headers="d54e16509" class="stentry" rowspan="1" colspan="1">filename</td>
                              <td valign="top" headers="d54e16512" class="stentry" rowspan="1" colspan="1">Path to the report with unresolved symbols.</td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e16503" class="stentry" rowspan="1" colspan="1">-o</td>
                              <td valign="top" headers="d54e16506" class="stentry" rowspan="1" colspan="1">--output</td>
                              <td valign="top" headers="d54e16509" class="stentry" rowspan="1" colspan="1">filename</td>
                              <td valign="top" headers="d54e16512" class="stentry" rowspan="1" colspan="1">Path and name of the output file. If it is omitted then
                                 "resolved" suffix is added to the original filename.
                              </td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e16503" class="stentry" rowspan="1" colspan="1">-d</td>
                              <td valign="top" headers="d54e16506" class="stentry" rowspan="1" colspan="1">--directories</td>
                              <td valign="top" headers="d54e16509" class="stentry" rowspan="1" colspan="1">directory paths</td>
                              <td valign="top" headers="d54e16512" class="stentry" rowspan="1" colspan="1">List of symbol folder paths, separated by semi-colon
                                 characters. Available only on Windows.
                              </td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e16503" class="stentry" rowspan="1" colspan="1">-v</td>
                              <td valign="top" headers="d54e16506" class="stentry" rowspan="1" colspan="1">--servers</td>
                              <td valign="top" headers="d54e16509" class="stentry" rowspan="1" colspan="1">server URLs</td>
                              <td valign="top" headers="d54e16512" class="stentry" rowspan="1" colspan="1">List of symbol servers that uses the same format as
                                 <samp class="ph codeph">_NT_SYMBOL_PATH</samp> environment variable,
                                 i.e. <samp class="ph codeph">srv*&lt;LocalStore&gt;*&lt;SymbolServerURL&gt;</samp>.
                                 Available only on Windows. 
                              </td>
                           </tr>
                           <tr class="strow">
                              <td valign="top" headers="d54e16503" class="stentry" rowspan="1" colspan="1">-n</td>
                              <td valign="top" headers="d54e16506" class="stentry" rowspan="1" colspan="1">--ignore-nt-sym-path</td>
                              <td valign="top" headers="d54e16509" class="stentry" rowspan="1" colspan="1">&nbsp;</td>
                              <td valign="top" headers="d54e16512" class="stentry" rowspan="1" colspan="1">Ignore the symbol locations stored in the
                                 <samp class="ph codeph">_NT_SYMBOL_PATH</samp> environment variable.
                                 Available only on Windows.
                              </td>
                           </tr>
                        </table>
                        <p class="p"></p>
                     </div>
                     <div class="topic nested2" id="broken-backtraces"><a name="broken-backtraces" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#broken-backtraces" name="broken-backtraces" shape="rect">Broken Backtraces on Tegra</a></h3>
                        <div class="body">
                           <p class="p">
                              In <span class="ph">Nsight Systems Embedded Platforms Edition</span>, in the symbols table there is a special
                              entry called <strong class="ph b">Broken backtraces</strong>. This entry is used to denote the
                              point in the call chain where the unwinding algorithms used by
                              <span class="ph">Nsight Systems</span> could not determine what is the next (caller)
                              function.
                           </p>
                           <p class="p">
                              Broken backtraces happen because there is no information related to the
                              current function that the unwinding algorithms can use. In the Top-Down
                              view, these functions are immediate children of the Broken backtraces row.
                           </p>
                           <p class="p">
                              One can eliminate broken backtraces by modifying the build system to provide
                              at least one kind of unwind information. The types of unwind information,
                              used by the algorithms in <span class="ph">Nsight Systems</span>, include the
                              following:
                           </p>
                           <p class="p">For ARMv7 binaries:</p>
                           <ul class="ul">
                              <li class="li">
                                 <p class="p">DWARF information in ELF sections: <samp class="ph codeph">.debug_frame</samp>,
                                    <samp class="ph codeph">.zdebug_frame</samp>, <samp class="ph codeph">.eh_frame</samp>,
                                    <samp class="ph codeph">.eh_frame_hdr</samp>. This information is the most precise.
                                    <samp class="ph codeph">.zdebug_frame</samp> is a compressed version of
                                    <samp class="ph codeph">.debug_frame</samp>, so at most one of them is typically present.
                                    <samp class="ph codeph">.eh_frame_hdr</samp> is a companion section for
                                    <samp class="ph codeph">.eh_frame</samp> and might be absent.
                                 </p>
                                 <p class="p">Compiler flag: <samp class="ph codeph">-g</samp>.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">Exception handling information in EHABI format provided in
                                    <samp class="ph codeph">.ARM.exidx</samp> and <samp class="ph codeph">.ARM.extab</samp> ELF sections.
                                    <samp class="ph codeph">.ARM.extab</samp> might be absent if all information is compact
                                    enough to be encoded into <samp class="ph codeph">.ARM.exidx</samp>.
                                 </p>
                                 <p class="p">Compiler flag: <samp class="ph codeph">-funwind-tables</samp>.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">Frame pointers (built into the <samp class="ph codeph">.text</samp> section).
                                 </p>
                                 <p class="p">Compiler flag: <samp class="ph codeph">-fno-omit-frame-pointer</samp>.
                                 </p>
                              </li>
                           </ul>
                           <p class="p">For Aarch64 binaries:</p>
                           <ul class="ul">
                              <li class="li">
                                 <p class="p">DWARF information in ELF sections: <samp class="ph codeph">.debug_frame</samp>,
                                    <samp class="ph codeph">.zdebug_frame</samp>, <samp class="ph codeph">.eh_frame</samp>,
                                    <samp class="ph codeph">.eh_frame_hdr</samp>. See additional comments above.
                                 </p>
                                 <p class="p">Compiler flag: <samp class="ph codeph">-g</samp>.
                                 </p>
                              </li>
                              <li class="li">
                                 <p class="p">Frame pointers (built into the <samp class="ph codeph">.text</samp> section).
                                 </p>
                                 <p class="p">Compiler flag: <samp class="ph codeph">-fno-omit-frame-pointer</samp>.
                                 </p>
                              </li>
                           </ul>
                           <p class="p">The following ELF sections should be considered empty if they have size of 4
                              bytes: <samp class="ph codeph">.debug_frame</samp>, <samp class="ph codeph">.eh_frame</samp>,
                              <samp class="ph codeph">.ARM.exidx</samp>. In this case, these sections only contain
                              termination records and no useful information.
                           </p>
                           <p class="p">For GCC, use the following compiler invocation to see which compiler flags
                              are enabled in your toolchain by default (for example, to check if
                              <samp class="ph codeph">-funwind-tables</samp> is enabled by default):
                           </p><pre xml:space="preserve">$ gcc -Q --help=common</pre><p class="p">
                              For GCC and Clang, add <samp class="ph codeph">-###</samp> to the compiler invocation
                              command to see which compiler flags are actually being used.
                           </p>
                           <p class="p">Since EHABI and DWARF information is compiled on per-unit basis (every
                              <samp class="ph codeph">.cpp</samp> or <samp class="ph codeph">.c</samp> file, as well as every static
                              library, can be built with or without this information), presence of the ELF
                              sections does not guarantee that every function has necessary unwind
                              information.
                           </p>
                           <p class="p">Frame pointers are required by the Aarch64 Procedure Call Standard. Adding
                              frame pointers slows down execution time, but in most cases the difference
                              is negligible.
                           </p>
                           <p class="p"></p>
                           <p class="p hr"></p>
                           <p class="p"></p>
                        </div>
                     </div>
                     <div class="topic nested2" id="elf-debug"><a name="elf-debug" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#elf-debug" name="elf-debug" shape="rect">Debug Versions of ELF Files</a></h3>
                        <div class="body">
                           <p class="p">
                              Often, after a binary is built, especially if it is built with debug
                              information (<samp class="ph codeph">-g</samp> compiler flag), it gets stripped before
                              deploying or installing.  In this case, ELF sections that contain useful
                              information, such as non-export function names or unwind information, can
                              get stripped as well.
                           </p>
                           <p class="p">
                              One solution is to deploy or install the original unstripped library instead
                              of the stripped one, but in many cases this would be inconvenient.
                              <span class="ph">Nsight Systems</span> can use missing information from alternative
                              locations.
                           </p>
                           <p class="p">
                              For target devices with Ubuntu, see <a class="xref" href="https://wiki.ubuntu.com/Debug_Symbol_packages" target="_blank" shape="rect">Debug Symbol Packages</a>. These packages typically
                              install debug ELF files with <samp class="ph codeph">/usr/lib/debug</samp> prefix.
                              <span class="ph">Nsight Systems</span> can find debug libraries there, and if it
                              matches the original library (e.g., the built-in <samp class="ph codeph">BuildID</samp> is
                              the same), it will be picked up and used to provide symbol names and unwind
                              information.
                           </p>
                           <p class="p">
                              Many packages have debug companions in the same repository and can be
                              directly installed with APT (<samp class="ph codeph">apt-get</samp>). Look for packages
                              with the <samp class="ph codeph">-dbg</samp> suffix. For other packages, refer to the 
                              <a class="xref" href="https://wiki.ubuntu.com/Debug_Symbol_packages" target="_blank" shape="rect">Debug Symbol Packages</a> wiki page on how to add the
                              debs package repository. After setting up the repository and running apt-get
                              update, look for packages with <samp class="ph codeph">-dbgsym</samp> suffix.
                           </p>
                           <p class="p">
                              To verify that a debug version of a library has been picked up and
                              downloaded from the target device, look in the <strong class="ph b">Module Summary</strong> section
                              of <strong class="ph b">Analysis Summary</strong>:
                           </p>
                           <blockquote class="lq">
                              <p class="p"><img class="image" src="graphics/debug-library-used.png" alt="Debug library has been used"></img></p>
                           </blockquote>
                           <p class="p"></p>
                           <p class="p hr"></p>
                           <p class="p"></p>
                        </div>
                     </div>
                  </div>
                  <div class="topic nested1" id="logging-troubleshooting"><a name="logging-troubleshooting" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#logging-troubleshooting" name="logging-troubleshooting" shape="rect">27.6.&nbsp;Logging</a></h3>
                     <div class="body">
                        <div class="p">To enable logging on the host, refer to this config file:
                           <pre xml:space="preserve">host-linux-x64/nvlog.config.template</pre></div>
                        <p class="p">When reporting any bugs please include the build version number as
                           described in the <strong class="ph b">Help → About</strong> dialog.  If possible, attach log
                           files and report (<samp class="ph codeph">.nsys-rep</samp>) files, as they already contain
                           necessary version information.
                        </p>
                     </div>
                     <div class="topic nested2" id="verbose-linux"><a name="verbose-linux" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#verbose-linux" name="verbose-linux" shape="rect">Verbose Logging on Linux Targets</a></h3>
                        <div class="body">
                           <p class="p">
                              Verbose logging is available when connecting to a Linux-based device from
                              the GUI on the host. This extra debug information is not available when
                              launching via the command line. <span class="ph">Nsight Systems</span> installs
                              its executable and library files into the following directory:
                           </p><pre xml:space="preserve">/opt/nvidia/nsight_systems/</pre><div class="p">
                              To enable verbose logging on the target device, when launched from the
                              host, follow these steps:
                              
                              <ol class="ol">
                                 <li class="li">
                                    <p class="p">Close the host application.</p>
                                 </li>
                                 <li class="li">
                                    <p class="p">Restart the target device.</p>
                                 </li>
                                 <li class="li">
                                    <p class="p">Place <samp class="ph codeph">nvlog.config</samp> from host directory to the <samp class="ph codeph">/opt/nvidia/nsight_systems</samp>
                                       directory on target.
                                    </p>
                                 </li>
                                 <li class="li">
                                    <p class="p">From SSH console, launch the following command:</p><pre xml:space="preserve">sudo /opt/nvidia/nsight_systems/nsys --daemon --debug</pre></li>
                                 <li class="li">
                                    <p class="p">Start the host application and connect to the target device.</p>
                                 </li>
                              </ol>
                           </div>
                           <p class="p">
                              Logs on the target devices are collected into this file (if enabled):
                           </p><pre xml:space="preserve">nsys.log</pre><p class="p">in the directory where <samp class="ph codeph">nsys</samp> command was launched.
                           </p>
                           <p class="p">Please note that in some cases, debug logging can significantly slow down
                              the profiler.
                           </p>
                        </div>
                     </div>
                     <div class="topic nested2" id="verbose-logging-windows"><a name="verbose-logging-windows" shape="rect">
                           <!-- --></a><h3 class="title topictitle2"><a href="#verbose-logging-windows" name="verbose-logging-windows" shape="rect">Verbose Logging on Windows Targets</a></h3>
                        <div class="body">
                           <p class="p">
                              Verbose logging is available when connecting to a Windows-based device
                              from the GUI on the host. <span class="ph">Nsight Systems</span> installs its
                              executable and library files into the following directory by default:
                           </p><pre xml:space="preserve">C:\Program Files\NVIDIA Corporation\<span class="ph">Nsight Systems</span>&nbsp;<span class="ph">2022.3</span></pre><div class="p">
                              To enable verbose logging on the target device, when launched from the
                              host, follow these steps:
                              
                              <ol class="ol">
                                 <li class="li">
                                    <p class="p">Close the host application.</p>
                                 </li>
                                 <li class="li">
                                    <p class="p">Terminate the <samp class="ph codeph">nsys</samp> process.
                                    </p>
                                 </li>
                                 <li class="li">
                                    <p class="p">Place <samp class="ph codeph">nvlog.config</samp> from host directory next to
                                       <span class="ph">Nsight Systems</span> Windows agent on the target device
                                    </p>
                                    <ul class="ul">
                                       <li class="li">
                                          <p class="p">Local Windows target:</p><pre xml:space="preserve">C:\Program Files\NVIDIA Corporation\<span class="ph">Nsight Systems</span>&nbsp;<span class="ph">2022.3</span>\target-windows-x64</pre></li>
                                       <li class="li">
                                          <p class="p">Remote Windows target:</p><pre xml:space="preserve">C:\Users\&lt;user name&gt;\AppData\Local\Temp\nvidia\nsight_systems</pre></li>
                                    </ul>
                                 </li>
                                 <li class="li">
                                    <p class="p">Start the host application and connect to the target device.</p>
                                 </li>
                              </ol>
                           </div>
                           <p class="p">Logs on the target devices are collected into this file (if enabled):</p><pre xml:space="preserve">nsight-sys.log</pre><p class="p">in the same directory as <span class="ph">Nsight Systems</span> Windows agent.
                           </p>
                           <p class="p">Please note that in some cases debug logging can significantly slow
                              down the profiler.
                           </p>
                        </div>
                     </div>
                  </div>
               </div>
               <div class="topic nested0" id="other_resources"><a name="other_resources" shape="rect">
                     <!-- --></a><h2 class="title topictitle1"><a href="#other_resources" name="other_resources" shape="rect">Other Resources</a></h2>
                  <div class="body">
                     <p class="p">Looking for information to help you use <span class="ph">Nsight Systems</span> the
                        most effectively? Here are some more resources you might want to review:
                     </p>
                  </div>
                  <div class="topic nested1" id="training_seminars"><a name="training_seminars" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#training_seminars" name="training_seminars" shape="rect">Training Seminars</a></h3>
                     <div class="body">
                        <p class="p">NVIDIA Deep Learning Institute Training - Self-Paced Online Course
                           <a class="xref" href="https://courses.nvidia.com/courses/course-v1:DLI+S-AC-03+V1/" target="_blank" shape="rect">Optimizing CUDA Machine Learning Codes With Nsight Profiling Tools</a></p>
                        <p class="p">
                           2018 NCSA Blue Waters Webinar - Video Only
                           <a class="xref" href="https://www.youtube.com/watch?v=WA8C48FJi3c" target="_blank" shape="rect">Introduction to NVIDIA Nsight Systems</a></p>
                     </div>
                  </div>
                  <div class="topic nested1" id="blog_posts"><a name="blog_posts" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#blog_posts" name="blog_posts" shape="rect">Blog Posts</a></h3>
                     <div class="body">
                        <p class="p">
                           NVIDIA developer blogs, these are longer form, technical pieces
                           written by tool and domain experts.
                        </p>
                        <div class="p">
                           <ul class="ul">
                              <li class="li">
                                 2021 - 
                                 <a class="xref" href="https://developer.nvidia.com/blog/optimizing-dx12-resource-uploads-to-the-gpu-using-cpu-visible-vram/" target="_blank" shape="rect">Optimizing DX12 Resource Uploads to the
                                    GPU Using CPU-Visible VRAM</a></li>
                              <li class="li">
                                 2019 - 
                                 <a class="xref" href="https://devblogs.nvidia.com/migrating-nvidia-nsight-tools-nvvp-nvprof/" target="_blank" shape="rect">Migrating to NVIDIA Nsight Tools from NVVP
                                    and nvprof</a></li>
                              <li class="li">
                                 2019 -
                                 <a class="xref" href="https://devblogs.nvidia.com/transitioning-nsight-systems-nvidia-visual-profiler-nvprof/" target="_blank" shape="rect">Transitioning to Nsight Systems from NVIDIA
                                    Visual Profiler / nvprof</a></li>
                              <li class="li">
                                 2019 -
                                 <a class="xref" href="https://devblogs.nvidia.com/nvidia-nsight-systems-adds-vulkan-support/" target="_blank" shape="rect">NVIDIA Nsight Systems Add Vulkan Support</a></li>
                              <li class="li">
                                 2019 -
                                 <a class="xref" href="https://devblogs.nvidia.com/tensorflow-performance-logging-plugin-nvtx-plugins-tf-public/" target="_blank" shape="rect">TensorFlow Performance Logging Plugin
                                    nvtx-plugins-tf Goes Public</a></li>
                              <li class="li">
                                 2020 -
                                 <a class="xref" href="https://developer.nvidia.com/blog/nvidia-nsight-systems-containers-cloud/" target="_blank" shape="rect">NVIDIA Nsight Systems in Containers and
                                    the Cloud</a></li>
                              <li class="li">
                                 2020 - 
                                 <a class="xref" href="https://developer.nvidia.com/blog/understanding-the-visualization-of-overhead-and-latency-in-nsight-systems/" target="_blank" shape="rect">Understanding the Visualization of Overhead and Latency in Nsight Systems</a></li>
                              <li class="li">
                                 2021 -
                                 <a class="xref" href="https://developer.nvidia.com/blog/optimizing-dx12-resource-uploads-to-the-gpu-using-cpu-visible-vram/" target="_blank" shape="rect">Optimizing DX12 Resource Uploads to the GPU Using CPU-Visible VRAM</a></li>
                           </ul>
                        </div>
                     </div>
                  </div>
                  <div class="topic nested1" id="feature_videos"><a name="feature_videos" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#feature_videos" name="feature_videos" shape="rect">Feature Videos</a></h3>
                     <div class="body">
                        <p class="p">Short videos, only a minute or two, to introduce new features.</p>
                        <div class="p">
                           <ul class="ul">
                              <li class="li"><a class="xref" href="https://youtu.be/ZeuM2k_hrq0" target="_blank" shape="rect">OpenMP Trace Feature Spotlight</a></li>
                              <li class="li"><a class="xref" href="https://youtu.be/r2ewwd4d0vc" target="_blank" shape="rect">Command Line Sessions Video Spotlight
                                    </a></li>
                              <li class="li"><a class="xref" href="https://youtu.be/DUhzjyBr-wg" target="_blank" shape="rect">Direct3D11 Feature Spotlight</a></li>
                              <li class="li"><a class="xref" href="https://youtu.be/witzRF-wu8M" target="_blank" shape="rect">Vulkan Trace</a></li>
                              <li class="li"><a class="xref" href="https://www.youtube.com/watch?v=fyhPFTF75tk" target="_blank" shape="rect">Statistics Driven Profiling</a></li>
                              <li class="li"><a class="xref" href="https://www.youtube.com/embed/bda5Er27jqo" target="_blank" shape="rect">Analyzing NCCL Usage with NVDIA Nsight
                                    Systems</a></li>
                           </ul>
                        </div>
                     </div>
                  </div>
                  <div class="topic nested1" id="conference_presentations"><a name="conference_presentations" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#conference_presentations" name="conference_presentations" shape="rect">Conference Presentations</a></h3>
                     <div class="body">
                        <div class="p">
                           <ul class="ul">
                              <li class="li">
                                 GTC 2022 - 
                                 <a class="xref" href="https://www.nvidia.com/en-us/on-demand/session/gtcspring22-s41518/" target="_blank" shape="rect">Killing Cloud Monsters Has Never Been
                                    Smoother</a></li>
                              <li class="li">
                                 GTC 2022 - 
                                 <a class="xref" href="https://www.nvidia.com/en-us/on-demand/session/gtcspring22-s41500/" target="_blank" shape="rect">Optimizing Communication with Nsight Systems
                                    Network Profiling</a></li>
                              <li class="li">
                                 GTC 2021 - 
                                 <a class="xref" href="https://www.nvidia.com/en-us/on-demand/session/gtcspring21-s31566/" target="_blank" shape="rect">Tuning GPU Network and Memory Usage in
                                    Apache Spark</a></li>
                              <li class="li">
                                 GTC 2020 - 
                                 <a class="xref" href="https://developer.nvidia.com/gtc/2020/video/s21547" target="_blank" shape="rect">Rebalancing the Load: Profile-Guided
                                    Optimization of the NAMD Molecular Dynamics Program for Modern GPUs
                                    using Nsight Systems</a></li>
                              <li class="li">
                                 GTC 2020 - 
                                 <a class="xref" href="https://developer.nvidia.com/gtc/2020/video/s21351" target="_blank" shape="rect">Scaling the Transformer Model
                                    Implementation in PyTorch Across Multiple Nodes</a></li>
                              <li class="li">
                                 GTC 2019 - 
                                 <a class="xref" href="https://on-demand-gtc.gputechconf.com/gtcnew/sessionview.php?sessionName=s9503-using+nsight+tools+to+optimize+the+namd+molecular+dynamics+simulation+program" target="_blank" shape="rect">Using Nsight Tools to Optimize the NAMD
                                    Molecular Dynamics Simulation Program</a></li>
                              <li class="li">
                                 GTC 2019 -
                                 <a class="xref" href="https://on-demand-gtc.gputechconf.com/gtcnew/sessionview.php?sessionName=s9866-optimizing+facebook+ai+workloads+for+nvidia+gpus" target="_blank" shape="rect">Optimizing Facebook AI Workloads for
                                    NVIDIA GPUs</a></li>
                              <li class="li">
                                 GTC 2018 -
                                 <a class="xref" href="https://on-demand-gtc.gputechconf.com/gtcnew/sessionview.php?sessionName=s8718-optimizing+hpc+simulation+and+visualization+codes+using+the+nvidia+nsight+systems" target="_blank" shape="rect">Optimizing HPC Simulation and
                                    Visualization Codes Using NVIDIA Nsight Systems</a></li>
                              <li class="li">
                                 GTC 2018 - Israel -
                                 <a class="xref" href="https://on-demand-gtc.gputechconf.com/gtcnew/sessionview.php?sessionName=sil8105-boost+dnn+training+performance+using+nvidia+tools" target="_blank" shape="rect">Boost DNN Training Performance using
                                    NVIDIA Tools</a></li>
                              <li class="li">Siggraph 2018 -
                                 <a class="xref" href="http://on-demand.gputechconf.com/siggraph/2018/video/sig1805-aurelio-reis-yaki-tebeka-taming-the-beast.html" target="_blank" shape="rect">Taming the Beast; Using NVIDIA Tools to
                                    Unlock Hidden GPU Performance</a></li>
                           </ul>
                        </div>
                     </div>
                  </div>
                  <div class="topic nested1" id="further-support"><a name="further-support" shape="rect">
                        <!-- --></a><h3 class="title topictitle2"><a href="#further-support" name="further-support" shape="rect">For More Support</a></h3>
                     <div class="body">
                        <p class="p">
                           To file a bug report or to ask a question on the
                           <span class="ph">Nsight Systems</span> forums, you will need to register with the
                           NVIDIA Developer Program. See the
                           <a class="xref" href="https://www.nvidia.com/en-us/account/faq/" target="_blank" shape="rect">FAQ</a>. You do not need to register to read the
                           forums.
                        </p>
                        <p class="p">
                           After that, you can access <span class="ph">Nsight Systems</span>&nbsp;<a class="xref" href="https://forums.developer.nvidia.com/c/development-tools/nsight-systems/116" target="_blank" shape="rect">Forums</a> and the <a class="xref" href="https://developer.nvidia.com/nvidia_bug/add" target="_blank" shape="rect">NVIDIA Bug Tracking System</a>.
                        </p>
                        <p class="p">
                           To submit feedback directly from the GUI, go to <strong class="ph b">Help-&gt;Send Feedback</strong>
                           and fill out the form. Enter your email address if you would like to
                           hear back from the <span class="ph">Nsight Systems</span> team.
                        </p>
                        <blockquote class="lq">
                           <p class="p"><img class="image" src="graphics/feedback.png" alt="Feedback"></img></p>
                        </blockquote>
                        <p class="p"></p>
                        <p class="p hr"></p>
                        <p class="p"></p>
                     </div>
                  </div>
               </div>
               
               <hr id="contents-end"></hr>
               
            </article>
         </div>
      </div>
      <script language="JavaScript" type="text/javascript" charset="utf-8" src="../common/formatting/common.min.js"></script>
      <script language="JavaScript" type="text/javascript" charset="utf-8" src="../common/scripts/google-analytics/google-analytics-write.js"></script>
      <script language="JavaScript" type="text/javascript" charset="utf-8" src="../common/scripts/google-analytics/google-analytics-tracker.js"></script>
      <script type="text/javascript">_satellite.pageBottom();</script>
      <script type="text/javascript">var switchTo5x=true;</script><script type="text/javascript" src="http://w.sharethis.com/button/buttons.js"></script><script type="text/javascript">stLight.options({publisher: "998dc202-a267-4d8e-bce9-14debadb8d92", doNotHash: false, doNotCopy: false, hashAddressBar: false});</script></body>
</html>