replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[DAGCombiner] allow undefs when matching vector splats for fmul folds 

llvm-svn: 343942
This commit is contained in:
Sanjay Patel 2018-10-07 16:05:37 +00:00
parent fcb1061c13
commit ef76e27985
2 changed files with 5 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
View File

@ -11083,8 +11083,8 @@ SDValue DAGCombiner::visitFSUB(SDNode *N) {
SDValue DAGCombiner::visitFMUL(SDNode *N) {
SDValue N0 = N->getOperand(0);
SDValue N1 = N->getOperand(1);
ConstantFPSDNode *N0CFP = isConstOrConstSplatFP(N0);
ConstantFPSDNode *N1CFP = isConstOrConstSplatFP(N1);
ConstantFPSDNode *N0CFP = isConstOrConstSplatFP(N0, true);
ConstantFPSDNode *N1CFP = isConstOrConstSplatFP(N1, true);
EVT VT = N->getValueType(0);
SDLoc DL(N);
const TargetOptions &Options = DAG.getTarget().Options;

8
llvm/test/CodeGen/X86/fmul-combines.ll
View File

@ -23,8 +23,7 @@ define <4 x float> @fmul_zero_nsz_nnan(<4 x float> %x) nounwind {
define <4 x float> @fmul_zero_nsz_nnan_undef(<4 x float> %x) nounwind {
; CHECK-LABEL: fmul_zero_nsz_nnan_undef:
; CHECK: # %bb.0:
; CHECK-NEXT: xorps %xmm1, %xmm1
; CHECK-NEXT: mulps %xmm1, %xmm0
; CHECK-NEXT: xorps %xmm0, %xmm0
; CHECK-NEXT: retq
%r = fmul nsz nnan <4 x float> %x, <float 0.0, float 0.0, float 0.0, float undef>
ret <4 x float> %r
@ -53,7 +52,7 @@ define <4 x float> @fmul2_v4f32(<4 x float> %x) {
define <4 x float> @fmul2_v4f32_undef(<4 x float> %x) {
; CHECK-LABEL: fmul2_v4f32_undef:
; CHECK: # %bb.0:
; CHECK-NEXT: mulps {{.*}}(%rip), %xmm0
; CHECK-NEXT: addps %xmm0, %xmm0
; CHECK-NEXT: retq
%y = fmul <4 x float> %x, <float undef, float 2.0, float 2.0, float 2.0>
ret <4 x float> %y
@ -89,8 +88,7 @@ define <4 x float> @fmul0_v4f32(<4 x float> %x) #0 {
define <4 x float> @fmul0_v4f32_undef(<4 x float> %x) #0 {
; CHECK-LABEL: fmul0_v4f32_undef:
; CHECK: # %bb.0:
; CHECK-NEXT: xorps %xmm1, %xmm1
; CHECK-NEXT: mulps %xmm1, %xmm0
; CHECK-NEXT: xorps %xmm0, %xmm0
; CHECK-NEXT: retq
%y = fmul <4 x float> %x, <float undef, float 0.0, float undef, float 0.0>
ret <4 x float> %y