maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[DAGCombine] Add (sext_in_reg (zext x)) -> (sext x) combine 

Handle the case where a sign extension has ended up being split into separate stages (typically to get around vector legal ops) and a zext + sext_in_reg gets inserted.

Differential Revision: https://reviews.llvm.org/D27461

llvm-svn: 288842
This commit is contained in:
Simon Pilgrim 2016-12-06 19:09:37 +00:00
parent 9b1b2de348
commit dd6ca639d5
2 changed files with 9 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
View File

@ -7137,6 +7137,15 @@ SDValue DAGCombiner::visitSIGN_EXTEND_INREG(SDNode *N) {
return DAG.getNode(ISD::SIGN_EXTEND, SDLoc(N), VT, N00, N1);
}
// fold (sext_in_reg (zext x)) -> (sext x)
// iff we are extending the source sign bit.
if (N0.getOpcode() == ISD::ZERO_EXTEND) {
SDValue N00 = N0.getOperand(0);
if (N00.getScalarValueSizeInBits() == EVTBits &&
(!LegalOperations || TLI.isOperationLegal(ISD::SIGN_EXTEND, VT)))
return DAG.getNode(ISD::SIGN_EXTEND, SDLoc(N), VT, N00, N1);
}
// fold (sext_in_reg x) -> (zext_in_reg x) if the sign bit is known zero.
if (DAG.MaskedValueIsZero(N0, APInt::getBitsSet(VTBits, EVTBits-1, EVTBits)))
return DAG.getZeroExtendInReg(N0, SDLoc(N), EVT.getScalarType());

12
llvm/test/CodeGen/X86/combine-sext-in-reg.ll
View File

@ -30,24 +30,12 @@ define <4 x i64> @sextinreg_zext_sext_v16i8_4i64(<16 x i8> %a0) {
; SSE-NEXT: pmovsxbq %xmm0, %xmm2
; SSE-NEXT: psrld $16, %xmm0
; SSE-NEXT: pmovsxbq %xmm0, %xmm1
; SSE-NEXT: psllq $32, %xmm2
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
; SSE-NEXT: psrad $31, %xmm2
; SSE-NEXT: pblendw {{.*#+}} xmm2 = xmm0[0,1],xmm2[2,3],xmm0[4,5],xmm2[6,7]
; SSE-NEXT: psllq $32, %xmm1
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,3,3]
; SSE-NEXT: psrad $31, %xmm1
; SSE-NEXT: pblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5],xmm1[6,7]
; SSE-NEXT: movdqa %xmm2, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: sextinreg_zext_sext_v16i8_4i64:
; AVX: # BB#0:
; AVX-NEXT: vpmovsxbq %xmm0, %ymm0
; AVX-NEXT: vpsllq $32, %ymm0, %ymm0
; AVX-NEXT: vpsrad $31, %ymm0, %ymm1
; AVX-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
; AVX-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
; AVX-NEXT: retq
%1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
%2 = sext <4 x i8> %1 to <4 x i32>