llvm/test/CodeGen/AMDGPU/inline-attr.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals all --version 5
; RUN: opt -mtriple=amdgpu-amd-amdhsa -S -O3 %s | FileCheck --check-prefixes=GCN,NONANS %s

declare void @extern() #0

define float @foo(float %x) #0 {
; GCN-LABEL: define float @foo(
; GCN-SAME: float [[X:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
; GCN-NEXT:  [[ENTRY:.*:]]
; GCN-NEXT:    tail call void @extern()
; GCN-NEXT:    [[MUL:%.*]] = fmul float [[X]], 1.500000e+01
; GCN-NEXT:    ret float [[MUL]]
;
entry:
  call void @extern()
  %mul = fmul float %x, 1.500000e+01
  ret float %mul
}

define amdgpu_kernel void @caller(ptr addrspace(1) %p) #1 {
; GCN-LABEL: define amdgpu_kernel void @caller(
; GCN-SAME: ptr addrspace(1) nofree captures(none) [[P:%.*]]) local_unnamed_addr #[[ATTR1:[0-9]+]] {
; GCN-NEXT:  [[ENTRY:.*:]]
; GCN-NEXT:    [[LOAD:%.*]] = load float, ptr addrspace(1) [[P]], align 4, !amdgpu.noclobber [[META0:![0-9]+]]
; GCN-NEXT:    tail call void @extern()
; GCN-NEXT:    [[MUL_I:%.*]] = fmul float [[LOAD]], 1.500000e+01
; GCN-NEXT:    store float [[MUL_I]], ptr addrspace(1) [[P]], align 4
; GCN-NEXT:    ret void
;
entry:
  %load = load float, ptr addrspace(1) %p, align 4
  %call = call fast float @foo(float %load)
  store float %call, ptr addrspace(1) %p, align 4
  ret void
}

attributes #0 = { nounwind }
attributes #1 = { nounwind "less-precise-fpmad"="true" }

; NOINFS: attributes #[[ATTR0]] = { nounwind }
; NOINFS: attributes #[[ATTR1]] = { nounwind "less-precise-fpmad"="false" }
; NOINFS: [[META0]] = !{}
;.
; GCN: attributes #[[ATTR0]] = { nounwind }
; GCN: attributes #[[ATTR1]] = { nounwind "less-precise-fpmad"="false" }
;.
; GCN: [[META0]] = !{}
;.
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; NONANS: {{.*}}