RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-23 11:13:28 +01:00
Code Issues Projects Releases Wiki Activity

[X86][AVX] Fold CONCAT(HOP(X,Y),HOP(Z,W)) -> HOP(CONCAT(X,Z),CONCAT(Y,W)) for integer types

Browse Source
This commit is contained in:
Simon Pilgrim 2020-08-05 14:30:26 +01:00
parent f2d8b0a536
commit 82176de7df
2 changed files with 7 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
lib/Target/X86/X86ISelLowering.cpp
View File

@ -48146,10 +48146,11 @@ static SDValue combineConcatVectorOps(const SDLoc &DL, MVT VT,
DAG.getNode(ISD::CONCAT_VECTORS, DL, SrcVT, RHS));
}
break;
case X86ISD::HADD:
case X86ISD::HSUB:
case X86ISD::PACKSS:
case X86ISD::PACKUS:
if (!IsSplat && NumOps == 2 && VT.is256BitVector() &&
Subtarget.hasInt256()) {
if (!IsSplat && VT.is256BitVector() && Subtarget.hasInt256()) {
SmallVector<SDValue, 2> LHS, RHS;
for (unsigned i = 0; i != NumOps; ++i) {
LHS.push_back(Ops[i].getOperand(0));

12
test/CodeGen/X86/haddsub-2.ll
View File

@ -579,11 +579,9 @@ define <8 x i32> @avx2_vphadd_d_test(<8 x i32> %A, <8 x i32> %B) {
;
; AVX2-LABEL: avx2_vphadd_d_test:
; AVX2: # %bb.0:
; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-NEXT: vphaddd %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-NEXT: vphaddd %xmm2, %xmm0, %xmm0
; AVX2-NEXT: vperm2i128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
; AVX2-NEXT: vphaddd %ymm2, %ymm0, %ymm0
; AVX2-NEXT: retq
%vecext = extractelement <8 x i32> %A, i32 0
%vecext1 = extractelement <8 x i32> %A, i32 1
@ -737,11 +735,9 @@ define <16 x i16> @avx2_vphadd_w_test(<16 x i16> %a, <16 x i16> %b) nounwind {
;
; AVX2-LABEL: avx2_vphadd_w_test:
; AVX2: # %bb.0:
; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
; AVX2-NEXT: vphaddw %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm2
; AVX2-NEXT: vphaddw %xmm2, %xmm0, %xmm0
; AVX2-NEXT: vperm2i128 {{.*#+}} ymm2 = ymm0[2,3],ymm1[2,3]
; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
; AVX2-NEXT: vphaddw %ymm2, %ymm0, %ymm0
; AVX2-NEXT: retq
%vecext = extractelement <16 x i16> %a, i32 0
%vecext1 = extractelement <16 x i16> %a, i32 1