RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-02-09 22:04:10 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

AMDGPU: Turn class x, p_zero|n_zero into fcmp oeq x, 0

Browse Source 
The library does use this for some reason.

llvm-svn: 339461
This commit is contained in:
Matt Arsenault 2018-08-10 18:58:49 +00:00
parent bf1675a1c4
commit 54e5db21e4
2 changed files with 16 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
lib/Transforms/InstCombine/InstCombineCalls.cpp
View File

@ -3195,6 +3195,15 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) {
return replaceInstUsesWith(*II, FCmp);
}
if (Mask == (N_ZERO | P_ZERO)) {
// Equivalent of == 0.
Value *FCmp = Builder.CreateFCmpOEQ(
Src0, ConstantFP::get(Src0->getType(), 0.0));
FCmp->takeName(II);
return replaceInstUsesWith(*II, FCmp);
}
const ConstantFP *CVal = dyn_cast<ConstantFP>(Src0);
if (!CVal) {
if (isa<UndefValue>(Src0))

7
test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
View File

@ -457,6 +457,13 @@ define i1 @test_class_isnan_f32(float %x) nounwind {
ret i1 %val
}
; CHECK-LABEL: @test_class_is_p0_n0_f32(
; CHECK: %val = fcmp oeq float %x, 0.000000e+00
define i1 @test_class_is_p0_n0_f32(float %x) nounwind {
%val = call i1 @llvm.amdgcn.class.f32(float %x, i32 96)
ret i1 %val
}
; CHECK-LABEL: @test_constant_class_snan_test_snan_f64(
; CHECK: ret i1 true
define i1 @test_constant_class_snan_test_snan_f64() nounwind {