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

[X86][AVX512] Add support for lowering shuffles to VSHUFPD 

llvm-svn: 274520
This commit is contained in:
Simon Pilgrim 2016-07-04 20:41:24 +00:00
parent 5d16cd9d63
commit 3ad040909a
3 changed files with 8 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

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

@ -11778,6 +11778,11 @@ static SDValue lowerV8F64VectorShuffle(const SDLoc &DL, ArrayRef<int> Mask,
lowerVectorShuffleWithUNPCK(DL, MVT::v8f64, Mask, V1, V2, DAG)) lowerVectorShuffleWithUNPCK(DL, MVT::v8f64, Mask, V1, V2, DAG))
return Unpck; return Unpck;
// Check if the blend happens to exactly fit that of SHUFPD.
if (SDValue Op =
lowerVectorShuffleWithSHUFPD(DL, MVT::v8f64, Mask, V1, V2, DAG))
return Op;
return lowerVectorShuffleWithPERMV(DL, MVT::v8f64, Mask, V1, V2, DAG); return lowerVectorShuffleWithPERMV(DL, MVT::v8f64, Mask, V1, V2, DAG);
} }

3
llvm/test/CodeGen/X86/sse3-avx-addsub.ll
View File

@ -149,8 +149,7 @@ define <8 x double> @test6(<8 x double> %A, <8 x double> %B) {
; AVX512: # BB#0: ; AVX512: # BB#0:
; AVX512-NEXT: vaddpd %zmm1, %zmm0, %zmm2 ; AVX512-NEXT: vaddpd %zmm1, %zmm0, %zmm2
; AVX512-NEXT: vsubpd %zmm1, %zmm0, %zmm0 ; AVX512-NEXT: vsubpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vmovdqa64 {{.*#+}} zmm1 = [0,9,2,11,4,13,6,15] ; AVX512-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[0],zmm2[1],zmm0[2],zmm2[3],zmm0[4],zmm2[5],zmm0[6],zmm2[7]
; AVX512-NEXT: vpermt2pd %zmm2, %zmm1, %zmm0
; AVX512-NEXT: retq ; AVX512-NEXT: retq
%add = fadd <8 x double> %A, %B %add = fadd <8 x double> %A, %B
%sub = fsub <8 x double> %A, %B %sub = fsub <8 x double> %A, %B

8
llvm/test/CodeGen/X86/vector-shuffle-512-v8.ll
View File

@ -217,16 +217,12 @@ define <8 x double> @shuffle_v8f64_81a3c5e7(<8 x double> %a, <8 x double> %b) {
; ;
; AVX512F-LABEL: shuffle_v8f64_81a3c5e7: ; AVX512F-LABEL: shuffle_v8f64_81a3c5e7:
; AVX512F: # BB#0: ; AVX512F: # BB#0:
; AVX512F-NEXT: vmovdqa64 {{.*#+}} zmm2 = [0,9,2,11,4,13,6,15] ; AVX512F-NEXT: vshufpd {{.*#+}} zmm0 = zmm1[0],zmm0[1],zmm1[2],zmm0[3],zmm1[4],zmm0[5],zmm1[6],zmm0[7]
; AVX512F-NEXT: vpermt2pd %zmm0, %zmm2, %zmm1
; AVX512F-NEXT: vmovaps %zmm1, %zmm0
; AVX512F-NEXT: retq ; AVX512F-NEXT: retq
; ;
; AVX512F-32-LABEL: shuffle_v8f64_81a3c5e7: ; AVX512F-32-LABEL: shuffle_v8f64_81a3c5e7:
; AVX512F-32: # BB#0: ; AVX512F-32: # BB#0:
; AVX512F-32-NEXT: vmovdqa64 {{.*#+}} zmm2 = [0,0,9,0,2,0,11,0,4,0,13,0,6,0,15,0] ; AVX512F-32-NEXT: vshufpd {{.*#+}} zmm0 = zmm1[0],zmm0[1],zmm1[2],zmm0[3],zmm1[4],zmm0[5],zmm1[6],zmm0[7]
; AVX512F-32-NEXT: vpermt2pd %zmm0, %zmm2, %zmm1
; AVX512F-32-NEXT: vmovaps %zmm1, %zmm0
; AVX512F-32-NEXT: retl ; AVX512F-32-NEXT: retl
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 8, i32 1, i32 10, i32 3, i32 12, i32 5, i32 14, i32 7> %shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 8, i32 1, i32 10, i32 3, i32 12, i32 5, i32 14, i32 7>
ret <8 x double> %shuffle ret <8 x double> %shuffle