File: generic-to-nvvm.ll

package info (click to toggle)
llvm-toolchain-20 1%3A20.1.8-1~exp1
  • links: PTS, VCS
  • area: main
  • in suites: experimental
  • size: 2,111,388 kB
  • sloc: cpp: 7,438,767; ansic: 1,393,871; asm: 1,012,926; python: 241,728; f90: 86,635; objc: 75,411; lisp: 42,144; pascal: 17,286; sh: 10,027; ml: 5,082; perl: 4,730; awk: 3,523; makefile: 3,349; javascript: 2,251; xml: 892; fortran: 672
file content (26 lines) | stat: -rw-r--r-- 921 bytes parent folder | download | duplicates (2) 
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
 
; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_20 | FileCheck %s
; RUN: %if ptxas %{ llc < %s -mtriple=nvptx64 -mcpu=sm_20 | %ptxas-verify %}

target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v16:16:16-v32:32:32-v64:64:64-v128:128:128-n16:32:64"
target triple = "nvptx-nvidia-cuda"

; Ensure global variables in address space 0 are promoted to address space 1

; CHECK: .global .align 4 .u32 myglobal = 42;
@myglobal = internal global i32 42, align 4
; CHECK: .global .align 4 .u32 myconst = 420;
@myconst = internal constant i32 420, align 4


define ptx_kernel void @foo(ptr %a, ptr %b) {
; Expect one load -- @myconst isn't loaded from, because we know its value
; statically.
; CHECK: ld.global.u32
; CHECK: st.global.u32
; CHECK: st.global.u32
  %ld1 = load i32, ptr @myglobal
  %ld2 = load i32, ptr @myconst
  store i32 %ld1, ptr %a
  store i32 %ld2, ptr %b
  ret void
}