ca03045d7f
We do not need this in the attributor, because `ST.getWavesPerEU` accounts for both the waves-per-eu and flat-workgroup-size attributes. If the waves-per-eu values are not valid, it drops them. In the attributor, we only need to propagate the values without using intermediate flat workgroup size values. Fixes SWDEV-550257.
62 lines
2.9 KiB
LLVM
62 lines
2.9 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals all --version 5
|
|
; RUN: opt -mtriple=amdgcn-amd-amdhsa -S -O3 -enable-unsafe-fp-math %s | FileCheck --check-prefixes=GCN,UNSAFE %s
|
|
; RUN: opt -mtriple=amdgcn-amd-amdhsa -S -O3 -enable-no-nans-fp-math %s | FileCheck --check-prefixes=GCN,NONANS %s
|
|
; RUN: opt -mtriple=amdgcn-amd-amdhsa -S -O3 -enable-no-infs-fp-math %s | FileCheck --check-prefixes=GCN,NOINFS %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) 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 "uniform-work-group-size"="false" "unsafe-fp-math"="true"}
|
|
attributes #1 = { nounwind "less-precise-fpmad"="true" "no-infs-fp-math"="true" "no-nans-fp-math"="true" "unsafe-fp-math"="true" }
|
|
|
|
;.
|
|
; UNSAFE: attributes #[[ATTR0]] = { nounwind "uniform-work-group-size"="false" "unsafe-fp-math"="true" }
|
|
; UNSAFE: attributes #[[ATTR1]] = { nounwind "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "uniform-work-group-size"="false" "unsafe-fp-math"="true" }
|
|
;.
|
|
; NONANS: attributes #[[ATTR0]] = { nounwind "no-nans-fp-math"="true" "uniform-work-group-size"="false" "unsafe-fp-math"="true" }
|
|
; NONANS: attributes #[[ATTR1]] = { nounwind "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="true" "uniform-work-group-size"="false" "unsafe-fp-math"="true" }
|
|
;.
|
|
; NOINFS: attributes #[[ATTR0]] = { nounwind "no-infs-fp-math"="true" "uniform-work-group-size"="false" "unsafe-fp-math"="true" }
|
|
; NOINFS: attributes #[[ATTR1]] = { nounwind "less-precise-fpmad"="false" "no-infs-fp-math"="true" "no-nans-fp-math"="false" "uniform-work-group-size"="false" "unsafe-fp-math"="true" }
|
|
;.
|
|
; UNSAFE: [[META0]] = !{}
|
|
;.
|
|
; NONANS: [[META0]] = !{}
|
|
;.
|
|
; NOINFS: [[META0]] = !{}
|
|
;.
|
|
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
|
|
; NOINFS: {{.*}}
|
|
; NONANS: {{.*}}
|
|
; UNSAFE: {{.*}}
|