xref: /aosp_15_r20/external/llvm/test/CodeGen/AMDGPU/register-count-comments.ll (revision 9880d6810fe72a1726cb53787c6711e909410d58)
1; RUN: llc -march=amdgcn -verify-machineinstrs -asm-verbose < %s | FileCheck -check-prefix=SI %s
2; RUN: llc -march=amdgcn -mtriple=amdgcn-unknown-amdhsa -verify-machineinstrs -asm-verbose -mattr=-flat-for-global < %s | FileCheck -check-prefix=SI %s
3
4declare i32 @llvm.amdgcn.mbcnt.lo(i32, i32) #0
5declare i32 @llvm.amdgcn.mbcnt.hi(i32, i32) #0
6
7; SI-LABEL: {{^}}foo:
8; SI: .section	.AMDGPU.csdata
9; SI: ; Kernel info:
10; SI: ; NumSgprs: {{[0-9]+}}
11; SI: ; NumVgprs: {{[0-9]+}}
12define void @foo(i32 addrspace(1)* noalias %out, i32 addrspace(1)* %abase, i32 addrspace(1)* %bbase) nounwind {
13  %mbcnt.lo = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0);
14  %tid = call i32 @llvm.amdgcn.mbcnt.hi(i32 -1, i32 %mbcnt.lo)
15  %aptr = getelementptr i32, i32 addrspace(1)* %abase, i32 %tid
16  %bptr = getelementptr i32, i32 addrspace(1)* %bbase, i32 %tid
17  %outptr = getelementptr i32, i32 addrspace(1)* %out, i32 %tid
18  %a = load i32, i32 addrspace(1)* %aptr, align 4
19  %b = load i32, i32 addrspace(1)* %bptr, align 4
20  %result = add i32 %a, %b
21  store i32 %result, i32 addrspace(1)* %outptr, align 4
22  ret void
23}
24
25; SI-LABEL: {{^}}one_vgpr_used:
26; SI: NumVgprs: 1
27define void @one_vgpr_used(i32 addrspace(1)* %out, i32 %x) nounwind {
28  store i32 %x, i32 addrspace(1)* %out, align 4
29  ret void
30}
31