mirror of
https://github.com/RPCS3/llvm.git
synced 2024-12-28 23:43:50 +00:00
0ff4287fe2
IR for CUDA should use "nvptx[64]-nvidia-cuda", and IR for NV OpenCL should use "nvptx[64]-nvidia-nvcl" git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@184579 91177308-0d34-0410-b5e6-96231b3b80d8
40 lines
1.2 KiB
LLVM
40 lines
1.2 KiB
LLVM
; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s
|
|
|
|
target triple = "nvptx-nvidia-cuda"
|
|
|
|
; Function Attrs: nounwind
|
|
; CHECK: .entry foo
|
|
define void @foo(float* nocapture %a) #0 {
|
|
%val = load float* %a
|
|
%tan = tail call fastcc float @__nv_fast_tanf(float %val)
|
|
store float %tan, float* %a
|
|
ret void
|
|
}
|
|
|
|
; Function Attrs: nounwind readnone
|
|
declare float @llvm.nvvm.sin.approx.ftz.f(float) #1
|
|
|
|
; Function Attrs: nounwind readnone
|
|
declare float @llvm.nvvm.cos.approx.ftz.f(float) #1
|
|
|
|
; Function Attrs: nounwind readnone
|
|
declare float @llvm.nvvm.div.approx.ftz.f(float, float) #1
|
|
|
|
; Function Attrs: alwaysinline inlinehint nounwind readnone
|
|
; CHECK: .func (.param .b32 func_retval0) __nv_fast_tanf
|
|
define internal fastcc float @__nv_fast_tanf(float %a) #2 {
|
|
entry:
|
|
%0 = tail call float @llvm.nvvm.sin.approx.ftz.f(float %a)
|
|
%1 = tail call float @llvm.nvvm.cos.approx.ftz.f(float %a)
|
|
%2 = tail call float @llvm.nvvm.div.approx.ftz.f(float %0, float %1)
|
|
ret float %2
|
|
}
|
|
|
|
attributes #0 = { nounwind }
|
|
attributes #1 = { nounwind readnone }
|
|
attributes #2 = { alwaysinline inlinehint nounwind readnone }
|
|
|
|
!nvvm.annotations = !{!0}
|
|
|
|
!0 = metadata !{void (float*)* @foo, metadata !"kernel", i32 1}
|