RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-23 03:02:36 +01:00
Code Issues Projects Releases Wiki Activity

[X86] Recognize constant splats in LowerFCOPYSIGN.

Browse Source 
llvm-svn: 345484
This commit is contained in:
Craig Topper 2018-10-28 23:51:35 +00:00
parent 2fa4b11ce0
commit 0fddf00499
2 changed files with 5 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
lib/Target/X86/X86ISelLowering.cpp
View File

@ -18060,7 +18060,7 @@ static SDValue LowerFCOPYSIGN(SDValue Op, SelectionDAG &DAG) {
// TODO: If we had general constant folding for FP logic ops, this check
// wouldn't be necessary.
SDValue MagBits;
if (ConstantFPSDNode *Op0CN = dyn_cast<ConstantFPSDNode>(Mag)) {
if (ConstantFPSDNode *Op0CN = isConstOrConstSplatFP(Mag)) {
APFloat APF = Op0CN->getValueAPF();
APF.clearSign();
MagBits = DAG.getConstantFP(APF, dl, LogicVT);

10
test/CodeGen/X86/sse1-fcopysign.ll
View File

@ -71,18 +71,16 @@ define <4 x float> @v4f32_neg(<4 x float> %a, <4 x float> %b) nounwind {
define <4 x float> @v4f32_const_mag(<4 x float> %a, <4 x float> %b) nounwind {
; X86-LABEL: v4f32_const_mag:
; X86: # %bb.0:
; X86-NEXT: andps {{\.LCPI.*}}, %xmm1
; X86-NEXT: movaps {{.*#+}} xmm0 = [1,1,1,1]
; X86-NEXT: movaps %xmm1, %xmm0
; X86-NEXT: andps {{\.LCPI.*}}, %xmm0
; X86-NEXT: orps %xmm1, %xmm0
; X86-NEXT: orps {{\.LCPI.*}}, %xmm0
; X86-NEXT: retl
;
; X64-LABEL: v4f32_const_mag:
; X64: # %bb.0:
; X64-NEXT: andps {{.*}}(%rip), %xmm1
; X64-NEXT: movaps {{.*#+}} xmm0 = [1,1,1,1]
; X64-NEXT: movaps %xmm1, %xmm0
; X64-NEXT: andps {{.*}}(%rip), %xmm0
; X64-NEXT: orps %xmm1, %xmm0
; X64-NEXT: orps {{.*}}(%rip), %xmm0
; X64-NEXT: retq
%tmp = tail call <4 x float> @llvm.copysign.v4f32(<4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>, <4 x float> %b )
ret <4 x float> %tmp