1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
|
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
<title>llvm-exegesis - LLVM Machine Instruction Benchmark — LLVM 13 documentation</title>
<link rel="stylesheet" href="../_static/pygments.css" type="text/css" />
<link rel="stylesheet" href="../_static/llvm-theme.css" type="text/css" />
<script id="documentation_options" data-url_root="../" src="../_static/documentation_options.js"></script>
<script src="../_static/jquery.js"></script>
<script src="../_static/underscore.js"></script>
<script src="../_static/doctools.js"></script>
<link rel="index" title="Index" href="../genindex.html" />
<link rel="search" title="Search" href="../search.html" />
<link rel="next" title="llvm-locstats - calculate statistics on DWARF debug location" href="llvm-locstats.html" />
<link rel="prev" title="lit - LLVM Integrated Tester" href="lit.html" />
<style type="text/css">
table.right { float: right; margin-left: 20px; }
table.right td { border: 1px solid #ccc; }
</style>
</head><body>
<div class="logo">
<a href="../index.html">
<img src="../_static/logo.png"
alt="LLVM Logo" width="250" height="88"/></a>
</div>
<div class="related" role="navigation" aria-label="related navigation">
<h3>Navigation</h3>
<ul>
<li class="right" style="margin-right: 10px">
<a href="../genindex.html" title="General Index"
accesskey="I">index</a></li>
<li class="right" >
<a href="llvm-locstats.html" title="llvm-locstats - calculate statistics on DWARF debug location"
accesskey="N">next</a> |</li>
<li class="right" >
<a href="lit.html" title="lit - LLVM Integrated Tester"
accesskey="P">previous</a> |</li>
<li><a href="https://llvm.org/">LLVM Home</a> | </li>
<li><a href="../index.html">Documentation</a>»</li>
<li class="nav-item nav-item-1"><a href="../Reference.html" >Reference</a> »</li>
<li class="nav-item nav-item-2"><a href="index.html" accesskey="U">LLVM Command Guide</a> »</li>
<li class="nav-item nav-item-this"><a href="">llvm-exegesis - LLVM Machine Instruction Benchmark</a></li>
</ul>
</div>
<div class="sphinxsidebar" role="navigation" aria-label="main navigation">
<div class="sphinxsidebarwrapper">
<h3>Documentation</h3>
<ul class="want-points">
<li><a href="https://llvm.org/docs/GettingStartedTutorials.html">Getting Started/Tutorials</a></li>
<li><a href="https://llvm.org/docs/UserGuides.html">User Guides</a></li>
<li><a href="https://llvm.org/docs/Reference.html">Reference</a></li>
</ul>
<h3>Getting Involved</h3>
<ul class="want-points">
<li><a href="https://llvm.org/docs/Contributing.html">Contributing to LLVM</a></li>
<li><a href="https://llvm.org/docs/HowToSubmitABug.html">Submitting Bug Reports</a></li>
<li><a href="https://llvm.org/docs/GettingInvolved.html#mailing-lists">Mailing Lists</a></li>
<li><a href="https://llvm.org/docs/GettingInvolved.html#irc">IRC</a></li>
<li><a href="https://llvm.org/docs/GettingInvolved.html#meetups-and-social-events">Meetups and Social Events</a></li>
</ul>
<h3>Additional Links</h3>
<ul class="want-points">
<li><a href="https://llvm.org/docs/FAQ.html">FAQ</a></li>
<li><a href="https://llvm.org/docs/Lexicon.html">Glossary</a></li>
<li><a href="https://llvm.org/pubs">Publications</a></li>
<li><a href="https://github.com/llvm/llvm-project//">Github Repository</a></li>
</ul>
<div role="note" aria-label="source link">
<h3>This Page</h3>
<ul class="this-page-menu">
<li><a href="../_sources/CommandGuide/llvm-exegesis.rst.txt"
rel="nofollow">Show Source</a></li>
</ul>
</div>
<div id="searchbox" style="display: none" role="search">
<h3 id="searchlabel">Quick search</h3>
<div class="searchformwrapper">
<form class="search" action="../search.html" method="get">
<input type="text" name="q" aria-labelledby="searchlabel" />
<input type="submit" value="Go" />
</form>
</div>
</div>
<script>$('#searchbox').show(0);</script>
</div>
</div>
<div class="document">
<div class="documentwrapper">
<div class="bodywrapper">
<div class="body" role="main">
<div class="section" id="llvm-exegesis-llvm-machine-instruction-benchmark">
<h1>llvm-exegesis - LLVM Machine Instruction Benchmark<a class="headerlink" href="#llvm-exegesis-llvm-machine-instruction-benchmark" title="Permalink to this headline">¶</a></h1>
<div class="section" id="synopsis">
<h2>SYNOPSIS<a class="headerlink" href="#synopsis" title="Permalink to this headline">¶</a></h2>
<p><strong class="program">llvm-exegesis</strong> [<em>options</em>]</p>
</div>
<div class="section" id="description">
<h2>DESCRIPTION<a class="headerlink" href="#description" title="Permalink to this headline">¶</a></h2>
<p><strong class="program">llvm-exegesis</strong> is a benchmarking tool that uses information available
in LLVM to measure host machine instruction characteristics like latency,
throughput, or port decomposition.</p>
<p>Given an LLVM opcode name and a benchmarking mode, <strong class="program">llvm-exegesis</strong>
generates a code snippet that makes execution as serial (resp. as parallel) as
possible so that we can measure the latency (resp. inverse throughput/uop decomposition)
of the instruction.
The code snippet is jitted and executed on the host subtarget. The time taken
(resp. resource usage) is measured using hardware performance counters. The
result is printed out as YAML to the standard output.</p>
<p>The main goal of this tool is to automatically (in)validate the LLVM’s TableDef
scheduling models. To that end, we also provide analysis of the results.</p>
<p><strong class="program">llvm-exegesis</strong> can also benchmark arbitrary user-provided code
snippets.</p>
</div>
<div class="section" id="example-1-benchmarking-instructions">
<h2>EXAMPLE 1: benchmarking instructions<a class="headerlink" href="#example-1-benchmarking-instructions" title="Permalink to this headline">¶</a></h2>
<p>Assume you have an X86-64 machine. To measure the latency of a single
instruction, run:</p>
<div class="highlight-bash notranslate"><div class="highlight"><pre><span></span>$ llvm-exegesis -mode<span class="o">=</span>latency -opcode-name<span class="o">=</span>ADD64rr
</pre></div>
</div>
<p>Measuring the uop decomposition or inverse throughput of an instruction works similarly:</p>
<div class="highlight-bash notranslate"><div class="highlight"><pre><span></span>$ llvm-exegesis -mode<span class="o">=</span>uops -opcode-name<span class="o">=</span>ADD64rr
$ llvm-exegesis -mode<span class="o">=</span>inverse_throughput -opcode-name<span class="o">=</span>ADD64rr
</pre></div>
</div>
<p>The output is a YAML document (the default is to write to stdout, but you can
redirect the output to a file using <cite>-benchmarks-file</cite>):</p>
<div class="highlight-none notranslate"><div class="highlight"><pre><span></span>---
key:
opcode_name: ADD64rr
mode: latency
config: ''
cpu_name: haswell
llvm_triple: x86_64-unknown-linux-gnu
num_repetitions: 10000
measurements:
- { key: latency, value: 1.0058, debug_string: '' }
error: ''
info: 'explicit self cycles, selecting one aliasing configuration.
Snippet:
ADD64rr R8, R8, R10
'
...
</pre></div>
</div>
<p>To measure the latency of all instructions for the host architecture, run:</p>
<div class="highlight-bash notranslate"><div class="highlight"><pre><span></span>$ llvm-exegesis -mode<span class="o">=</span>latency -opcode-index<span class="o">=</span>-1
</pre></div>
</div>
</div>
<div class="section" id="example-2-benchmarking-a-custom-code-snippet">
<h2>EXAMPLE 2: benchmarking a custom code snippet<a class="headerlink" href="#example-2-benchmarking-a-custom-code-snippet" title="Permalink to this headline">¶</a></h2>
<p>To measure the latency/uops of a custom piece of code, you can specify the
<cite>snippets-file</cite> option (<cite>-</cite> reads from standard input).</p>
<div class="highlight-bash notranslate"><div class="highlight"><pre><span></span>$ <span class="nb">echo</span> <span class="s2">"vzeroupper"</span> <span class="p">|</span> llvm-exegesis -mode<span class="o">=</span>uops -snippets-file<span class="o">=</span>-
</pre></div>
</div>
<p>Real-life code snippets typically depend on registers or memory.
<strong class="program">llvm-exegesis</strong> checks the liveliness of registers (i.e. any register
use has a corresponding def or is a “live in”). If your code depends on the
value of some registers, you have two options:</p>
<ul class="simple">
<li><p>Mark the register as requiring a definition. <strong class="program">llvm-exegesis</strong> will
automatically assign a value to the register. This can be done using the
directive <cite>LLVM-EXEGESIS-DEFREG <reg name> <hex_value></cite>, where <cite><hex_value></cite>
is a bit pattern used to fill <cite><reg_name></cite>. If <cite><hex_value></cite> is smaller than
the register width, it will be sign-extended.</p></li>
<li><p>Mark the register as a “live in”. <strong class="program">llvm-exegesis</strong> will benchmark
using whatever value was in this registers on entry. This can be done using
the directive <cite>LLVM-EXEGESIS-LIVEIN <reg name></cite>.</p></li>
</ul>
<p>For example, the following code snippet depends on the values of XMM1 (which
will be set by the tool) and the memory buffer passed in RDI (live in).</p>
<div class="highlight-none notranslate"><div class="highlight"><pre><span></span># LLVM-EXEGESIS-LIVEIN RDI
# LLVM-EXEGESIS-DEFREG XMM1 42
vmulps (%rdi), %xmm1, %xmm2
vhaddps %xmm2, %xmm2, %xmm3
addq $0x10, %rdi
</pre></div>
</div>
</div>
<div class="section" id="example-3-analysis">
<h2>EXAMPLE 3: analysis<a class="headerlink" href="#example-3-analysis" title="Permalink to this headline">¶</a></h2>
<p>Assuming you have a set of benchmarked instructions (either latency or uops) as
YAML in file <cite>/tmp/benchmarks.yaml</cite>, you can analyze the results using the
following command:</p>
<div class="highlight-bash notranslate"><div class="highlight"><pre><span></span> $ llvm-exegesis -mode<span class="o">=</span>analysis <span class="se">\</span>
-benchmarks-file<span class="o">=</span>/tmp/benchmarks.yaml <span class="se">\</span>
-analysis-clusters-output-file<span class="o">=</span>/tmp/clusters.csv <span class="se">\</span>
-analysis-inconsistencies-output-file<span class="o">=</span>/tmp/inconsistencies.html
</pre></div>
</div>
<p>This will group the instructions into clusters with the same performance
characteristics. The clusters will be written out to <cite>/tmp/clusters.csv</cite> in the
following format:</p>
<div class="highlight-none notranslate"><div class="highlight"><pre><span></span>cluster_id,opcode_name,config,sched_class
...
2,ADD32ri8_DB,,WriteALU,1.00
2,ADD32ri_DB,,WriteALU,1.01
2,ADD32rr,,WriteALU,1.01
2,ADD32rr_DB,,WriteALU,1.00
2,ADD32rr_REV,,WriteALU,1.00
2,ADD64i32,,WriteALU,1.01
2,ADD64ri32,,WriteALU,1.01
2,MOVSX64rr32,,BSWAP32r_BSWAP64r_MOVSX64rr32,1.00
2,VPADDQYrr,,VPADDBYrr_VPADDDYrr_VPADDQYrr_VPADDWYrr_VPSUBBYrr_VPSUBDYrr_VPSUBQYrr_VPSUBWYrr,1.02
2,VPSUBQYrr,,VPADDBYrr_VPADDDYrr_VPADDQYrr_VPADDWYrr_VPSUBBYrr_VPSUBDYrr_VPSUBQYrr_VPSUBWYrr,1.01
2,ADD64ri8,,WriteALU,1.00
2,SETBr,,WriteSETCC,1.01
...
</pre></div>
</div>
<p><strong class="program">llvm-exegesis</strong> will also analyze the clusters to point out
inconsistencies in the scheduling information. The output is an html file. For
example, <cite>/tmp/inconsistencies.html</cite> will contain messages like the following :</p>
<img alt="../_images/llvm-exegesis-analysis.png" class="align-center" src="../_images/llvm-exegesis-analysis.png" />
<p>Note that the scheduling class names will be resolved only when
<strong class="program">llvm-exegesis</strong> is compiled in debug mode, else only the class id will
be shown. This does not invalidate any of the analysis results though.</p>
</div>
<div class="section" id="options">
<h2>OPTIONS<a class="headerlink" href="#options" title="Permalink to this headline">¶</a></h2>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-help">
<code class="sig-name descname"><span class="pre">-help</span></code><code class="sig-prename descclassname"></code><a class="headerlink" href="#cmdoption-llvm-exegesis-help" title="Permalink to this definition">¶</a></dt>
<dd><p>Print a summary of command line options.</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-opcode-index">
<code class="sig-name descname"><span class="pre">-opcode-index</span></code><code class="sig-prename descclassname"><span class="pre">=<LLVM</span> <span class="pre">opcode</span> <span class="pre">index></span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-opcode-index" title="Permalink to this definition">¶</a></dt>
<dd><p>Specify the opcode to measure, by index. Specifying <cite>-1</cite> will result
in measuring every existing opcode. See example 1 for details.
Either <cite>opcode-index</cite>, <cite>opcode-name</cite> or <cite>snippets-file</cite> must be set.</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-opcode-name">
<code class="sig-name descname"><span class="pre">-opcode-name</span></code><code class="sig-prename descclassname"><span class="pre">=<opcode</span> <span class="pre">name</span> <span class="pre">1>,<opcode</span> <span class="pre">name</span> <span class="pre">2>,...</span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-opcode-name" title="Permalink to this definition">¶</a></dt>
<dd><p>Specify the opcode to measure, by name. Several opcodes can be specified as
a comma-separated list. See example 1 for details.
Either <cite>opcode-index</cite>, <cite>opcode-name</cite> or <cite>snippets-file</cite> must be set.</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-snippets-file">
<code class="sig-name descname"><span class="pre">-snippets-file</span></code><code class="sig-prename descclassname"><span class="pre">=<filename></span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-snippets-file" title="Permalink to this definition">¶</a></dt>
<dd><p>Specify the custom code snippet to measure. See example 2 for details.
Either <cite>opcode-index</cite>, <cite>opcode-name</cite> or <cite>snippets-file</cite> must be set.</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-mode">
<code class="sig-name descname"><span class="pre">-mode</span></code><code class="sig-prename descclassname"><span class="pre">=[latency|uops|inverse_throughput|analysis]</span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-mode" title="Permalink to this definition">¶</a></dt>
<dd><p>Specify the run mode. Note that some modes have additional requirements and options.</p>
<p><cite>latency</cite> mode can be make use of either RDTSC or LBR.
<cite>latency[LBR]</cite> is only available on X86 (at least <cite>Skylake</cite>).
To run in <cite>latency</cite> mode, a positive value must be specified
for <cite>x86-lbr-sample-period</cite> and <cite>–repetition-mode=loop</cite>.</p>
<p>In <cite>analysis</cite> mode, you also need to specify at least one of the
<cite>-analysis-clusters-output-file=</cite> and <cite>-analysis-inconsistencies-output-file=</cite>.</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-x86-lbr-sample-period">
<code class="sig-name descname"><span class="pre">-x86-lbr-sample-period</span></code><code class="sig-prename descclassname"><span class="pre">=<nBranches/sample></span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-x86-lbr-sample-period" title="Permalink to this definition">¶</a></dt>
<dd><p>Specify the LBR sampling period - how many branches before we take a sample.
When a positive value is specified for this option and when the mode is <cite>latency</cite>,
we will use LBRs for measuring.
On choosing the “right” sampling period, a small value is preferred, but throttling
could occur if the sampling is too frequent. A prime number should be used to
avoid consistently skipping certain blocks.</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-repetition-mode">
<code class="sig-name descname"><span class="pre">-repetition-mode</span></code><code class="sig-prename descclassname"><span class="pre">=[duplicate|loop|min]</span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-repetition-mode" title="Permalink to this definition">¶</a></dt>
<dd><p>Specify the repetition mode. <cite>duplicate</cite> will create a large, straight line
basic block with <cite>num-repetitions</cite> instructions (repeating the snippet
<cite>num-repetitions</cite>/<cite>snippet size</cite> times). <cite>loop</cite> will, optionally, duplicate the
snippet until the loop body contains at least <cite>loop-body-size</cite> instructions,
and then wrap the result in a loop which will execute <cite>num-repetitions</cite>
instructions (thus, again, repeating the snippet
<cite>num-repetitions</cite>/<cite>snippet size</cite> times). The <cite>loop</cite> mode, especially with loop
unrolling tends to better hide the effects of the CPU frontend on architectures
that cache decoded instructions, but consumes a register for counting
iterations. If performing an analysis over many opcodes, it may be best to
instead use the <cite>min</cite> mode, which will run each other mode,
and produce the minimal measured result.</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-num-repetitions">
<code class="sig-name descname"><span class="pre">-num-repetitions</span></code><code class="sig-prename descclassname"><span class="pre">=<Number</span> <span class="pre">of</span> <span class="pre">repetitions></span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-num-repetitions" title="Permalink to this definition">¶</a></dt>
<dd><p>Specify the target number of executed instructions. Note that the actual
repetition count of the snippet will be <cite>num-repetitions</cite>/<cite>snippet size</cite>.
Higher values lead to more accurate measurements but lengthen the benchmark.</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-loop-body-size">
<code class="sig-name descname"><span class="pre">-loop-body-size</span></code><code class="sig-prename descclassname"><span class="pre">=<Preferred</span> <span class="pre">loop</span> <span class="pre">body</span> <span class="pre">size></span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-loop-body-size" title="Permalink to this definition">¶</a></dt>
<dd><p>Only effective for <cite>-repetition-mode=[loop|min]</cite>.
Instead of looping over the snippet directly, first duplicate it so that the
loop body contains at least this many instructions. This potentially results
in loop body being cached in the CPU Op Cache / Loop Cache, which allows to
which may have higher throughput than the CPU decoders.</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-max-configs-per-opcode">
<code class="sig-name descname"><span class="pre">-max-configs-per-opcode</span></code><code class="sig-prename descclassname"><span class="pre">=<value></span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-max-configs-per-opcode" title="Permalink to this definition">¶</a></dt>
<dd><p>Specify the maximum configurations that can be generated for each opcode.
By default this is <cite>1</cite>, meaning that we assume that a single measurement is
enough to characterize an opcode. This might not be true of all instructions:
for example, the performance characteristics of the LEA instruction on X86
depends on the value of assigned registers and immediates. Setting a value of
<cite>-max-configs-per-opcode</cite> larger than <cite>1</cite> allows <cite>llvm-exegesis</cite> to explore
more configurations to discover if some register or immediate assignments
lead to different performance characteristics.</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-benchmarks-file">
<code class="sig-name descname"><span class="pre">-benchmarks-file</span></code><code class="sig-prename descclassname"><span class="pre">=</path/to/file></span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-benchmarks-file" title="Permalink to this definition">¶</a></dt>
<dd><p>File to read (<cite>analysis</cite> mode) or write (<cite>latency</cite>/<cite>uops</cite>/<cite>inverse_throughput</cite>
modes) benchmark results. “-” uses stdin/stdout.</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-analysis-clusters-output-file">
<code class="sig-name descname"><span class="pre">-analysis-clusters-output-file</span></code><code class="sig-prename descclassname"><span class="pre">=</path/to/file></span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-analysis-clusters-output-file" title="Permalink to this definition">¶</a></dt>
<dd><p>If provided, write the analysis clusters as CSV to this file. “-” prints to
stdout. By default, this analysis is not run.</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-analysis-inconsistencies-output-file">
<code class="sig-name descname"><span class="pre">-analysis-inconsistencies-output-file</span></code><code class="sig-prename descclassname"><span class="pre">=</path/to/file></span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-analysis-inconsistencies-output-file" title="Permalink to this definition">¶</a></dt>
<dd><p>If non-empty, write inconsistencies found during analysis to this file. <cite>-</cite>
prints to stdout. By default, this analysis is not run.</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-analysis-clustering">
<code class="sig-name descname"><span class="pre">-analysis-clustering</span></code><code class="sig-prename descclassname"><span class="pre">=[dbscan,naive]</span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-analysis-clustering" title="Permalink to this definition">¶</a></dt>
<dd><p>Specify the clustering algorithm to use. By default DBSCAN will be used.
Naive clustering algorithm is better for doing further work on the
<cite>-analysis-inconsistencies-output-file=</cite> output, it will create one cluster
per opcode, and check that the cluster is stable (all points are neighbours).</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-analysis-numpoints">
<code class="sig-name descname"><span class="pre">-analysis-numpoints</span></code><code class="sig-prename descclassname"><span class="pre">=<dbscan</span> <span class="pre">numPoints</span> <span class="pre">parameter></span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-analysis-numpoints" title="Permalink to this definition">¶</a></dt>
<dd><p>Specify the numPoints parameters to be used for DBSCAN clustering
(<cite>analysis</cite> mode, DBSCAN only).</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-analysis-clustering-epsilon">
<code class="sig-name descname"><span class="pre">-analysis-clustering-epsilon</span></code><code class="sig-prename descclassname"><span class="pre">=<dbscan</span> <span class="pre">epsilon</span> <span class="pre">parameter></span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-analysis-clustering-epsilon" title="Permalink to this definition">¶</a></dt>
<dd><p>Specify the epsilon parameter used for clustering of benchmark points
(<cite>analysis</cite> mode).</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-analysis-inconsistency-epsilon">
<code class="sig-name descname"><span class="pre">-analysis-inconsistency-epsilon</span></code><code class="sig-prename descclassname"><span class="pre">=<epsilon></span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-analysis-inconsistency-epsilon" title="Permalink to this definition">¶</a></dt>
<dd><p>Specify the epsilon parameter used for detection of when the cluster
is different from the LLVM schedule profile values (<cite>analysis</cite> mode).</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-analysis-display-unstable-clusters">
<code class="sig-name descname"><span class="pre">-analysis-display-unstable-clusters</span></code><code class="sig-prename descclassname"></code><a class="headerlink" href="#cmdoption-llvm-exegesis-analysis-display-unstable-clusters" title="Permalink to this definition">¶</a></dt>
<dd><p>If there is more than one benchmark for an opcode, said benchmarks may end up
not being clustered into the same cluster if the measured performance
characteristics are different. by default all such opcodes are filtered out.
This flag will instead show only such unstable opcodes.</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-ignore-invalid-sched-class">
<code class="sig-name descname"><span class="pre">-ignore-invalid-sched-class</span></code><code class="sig-prename descclassname"><span class="pre">=false</span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-ignore-invalid-sched-class" title="Permalink to this definition">¶</a></dt>
<dd><p>If set, ignore instructions that do not have a sched class (class idx = 0).</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-mcpu">
<code class="sig-name descname"><span class="pre">-mcpu</span></code><code class="sig-prename descclassname"><span class="pre">=<cpu</span> <span class="pre">name></span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-mcpu" title="Permalink to this definition">¶</a></dt>
<dd><p>If set, measure the cpu characteristics using the counters for this CPU. This
is useful when creating new sched models (the host CPU is unknown to LLVM).</p>
</dd></dl>
<dl class="std option">
<dt id="cmdoption-llvm-exegesis-dump-object-to-disk">
<code class="sig-name descname"><span class="pre">--dump-object-to-disk</span></code><code class="sig-prename descclassname"><span class="pre">=true</span></code><a class="headerlink" href="#cmdoption-llvm-exegesis-dump-object-to-disk" title="Permalink to this definition">¶</a></dt>
<dd><p>By default, llvm-exegesis will dump the generated code to a temporary file to
enable code inspection. You may disable it to speed up the execution and save
disk space.</p>
</dd></dl>
</div>
<div class="section" id="exit-status">
<h2>EXIT STATUS<a class="headerlink" href="#exit-status" title="Permalink to this headline">¶</a></h2>
<p><strong class="program">llvm-exegesis</strong> returns 0 on success. Otherwise, an error message is
printed to standard error, and the tool returns a non 0 value.</p>
</div>
</div>
<div class="clearer"></div>
</div>
</div>
</div>
<div class="clearer"></div>
</div>
<div class="related" role="navigation" aria-label="related navigation">
<h3>Navigation</h3>
<ul>
<li class="right" style="margin-right: 10px">
<a href="../genindex.html" title="General Index"
>index</a></li>
<li class="right" >
<a href="llvm-locstats.html" title="llvm-locstats - calculate statistics on DWARF debug location"
>next</a> |</li>
<li class="right" >
<a href="lit.html" title="lit - LLVM Integrated Tester"
>previous</a> |</li>
<li><a href="https://llvm.org/">LLVM Home</a> | </li>
<li><a href="../index.html">Documentation</a>»</li>
<li class="nav-item nav-item-1"><a href="../Reference.html" >Reference</a> »</li>
<li class="nav-item nav-item-2"><a href="index.html" >LLVM Command Guide</a> »</li>
<li class="nav-item nav-item-this"><a href="">llvm-exegesis - LLVM Machine Instruction Benchmark</a></li>
</ul>
</div>
<div class="footer" role="contentinfo">
© Copyright 2003-2021, LLVM Project.
Last updated on 2021-09-18.
Created using <a href="https://www.sphinx-doc.org/">Sphinx</a> 3.5.4.
</div>
</body>
</html>
|
