llvm/test/CodeGen/AMDGPU/endcf-loop-header.ll
Matt Arsenault d706d030af AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel
Currently the default C calling convention functions are treated
the same as compute kernels. Make this explicit so the default
calling convention can be changed to a non-kernel.

Converted with perl -pi -e 's/define void/define amdgpu_kernel void/'
on the relevant test directories (and undoing in one place that actually
wanted a non-kernel).

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@298444 91177308-0d34-0410-b5e6-96231b3b80d8
2017-03-21 21:39:51 +00:00

40 lines
1.0 KiB
LLVM

; RUN: llc < %s -march=amdgcn -mcpu=SI -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:[0-9A-Za-z_]+]]: ; %loop{{$}}
; CHECK-NOT: s_or_b64 exec, exec
; CHECK: s_cbranch_execnz [[LOOP_LABEL]]
define amdgpu_kernel void @test(i32 addrspace(1)* %out) {
entry:
%cond = call i32 @llvm.r600.read.tidig.x() #0
%tmp0 = icmp eq i32 %cond, 0
br i1 %tmp0, label %if, label %loop
if:
store i32 0, i32 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, i32 addrspace(1)* %out, i64 1
store i32 %inc, i32 addrspace(1)* %tmp3
ret void
}
declare i32 @llvm.r600.read.tidig.x() #0
attributes #0 = { readnone }