File: endcf-loop-header.ll

package info (click to toggle)
llvm-toolchain-17 1%3A17.0.6-22
  • links: PTS, VCS
  • area: main
  • in suites: forky, sid, trixie
  • size: 1,799,624 kB
  • sloc: cpp: 6,428,607; ansic: 1,383,196; asm: 793,408; python: 223,504; objc: 75,364; f90: 60,502; lisp: 33,869; pascal: 15,282; sh: 9,684; perl: 7,453; ml: 4,937; awk: 3,523; makefile: 2,889; javascript: 2,149; xml: 888; fortran: 619; cs: 573
file content (39 lines) | stat: -rw-r--r-- 1,076 bytes parent folder | download | duplicates (4) 
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
 
; RUN: llc < %s -march=amdgcn -verify-machineinstrs | FileCheck %s

; This tests that the llvm.SI.end.cf intrinsic is not inserted into the
; loop block.  This intrinsic will be lowered to s_or_b64 by the code
; generator.

; CHECK-LABEL: {{^}}test:

; This is was lowered from the llvm.SI.end.cf intrinsic:
; CHECK: s_or_b64 exec, exec

; CHECK: [[LOOP_LABEL:.L[0-9A-Za-z_]+]]: ; %loop{{$}}
; CHECK-NOT: s_or_b64 exec, exec
; CHECK: s_cbranch_execnz [[LOOP_LABEL]]
define amdgpu_kernel void @test(ptr addrspace(1) %out) {
entry:
  %cond = call i32 @llvm.amdgcn.workitem.id.x() #0
  %tmp0 = icmp eq i32 %cond, 0
  br i1 %tmp0, label %if, label %loop

if:
  store i32 0, ptr addrspace(1) %out
  br label %loop

loop:
  %tmp1 = phi i32 [0, %entry], [0, %if], [%inc, %loop]
  %inc = add i32 %tmp1, %cond
  %tmp2 = icmp ugt i32 %inc, 10
  br i1 %tmp2, label %done, label %loop

done:
  %tmp3 = getelementptr i32, ptr addrspace(1) %out, i64 1
  store i32 %inc, ptr addrspace(1) %tmp3
  ret void
}

declare i32 @llvm.amdgcn.workitem.id.x() #0

attributes #0 = { nounwind readnone }