diff options
| author | Craig Topper <craig.topper@gmail.com> | 2016-09-05 06:43:06 +0000 |
|---|---|---|
| committer | Craig Topper <craig.topper@gmail.com> | 2016-09-05 06:43:06 +0000 |
| commit | d9ca3d97efa24cfd121a6f3beab7bf8af38d0cb7 (patch) | |
| tree | cdfb945eddfa37651647d06e2da05c5952583dca /llvm/test/CodeGen/X86/avx512-ext.ll | |
| parent | 87a675be507204a2b2cff32d960463d314b328ed (diff) | |
| download | bcm5719-llvm-d9ca3d97efa24cfd121a6f3beab7bf8af38d0cb7.tar.gz bcm5719-llvm-d9ca3d97efa24cfd121a6f3beab7bf8af38d0cb7.zip | |
[AVX-512] Simplify X86InstrInfo::copyPhysReg for 128/256-bit vectors with AVX512, but not VLX. We should use the VEX opcodes and trust the register allocator to not use the extended XMM/YMM register space.
Previously we were extending to copying the whole ZMM register. The register allocator shouldn't use XMM16-31 or YMM16-31 in this configuration as the instructions to spill them aren't available.
llvm-svn: 280648
Diffstat (limited to 'llvm/test/CodeGen/X86/avx512-ext.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/avx512-ext.ll | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/llvm/test/CodeGen/X86/avx512-ext.ll b/llvm/test/CodeGen/X86/avx512-ext.ll index e45dda1ec8f..4d8f95a00f5 100644 --- a/llvm/test/CodeGen/X86/avx512-ext.ll +++ b/llvm/test/CodeGen/X86/avx512-ext.ll @@ -163,7 +163,7 @@ define <32 x i16> @zext_32x8mem_to_32x16(<32 x i8> *%i , <32 x i1> %mask) nounwi ; KNL-NEXT: vpsllw $15, %ymm0, %ymm0 ; KNL-NEXT: vpsraw $15, %ymm0, %ymm0 ; KNL-NEXT: vpand %ymm1, %ymm0, %ymm1 -; KNL-NEXT: vmovdqa64 %zmm2, %zmm0 +; KNL-NEXT: vmovdqa %ymm2, %ymm0 ; KNL-NEXT: retq ; ; SKX-LABEL: zext_32x8mem_to_32x16: @@ -192,7 +192,7 @@ define <32 x i16> @sext_32x8mem_to_32x16(<32 x i8> *%i , <32 x i1> %mask) nounwi ; KNL-NEXT: vpsllw $15, %ymm0, %ymm0 ; KNL-NEXT: vpsraw $15, %ymm0, %ymm0 ; KNL-NEXT: vpand %ymm1, %ymm0, %ymm1 -; KNL-NEXT: vmovdqa64 %zmm2, %zmm0 +; KNL-NEXT: vmovdqa %ymm2, %ymm0 ; KNL-NEXT: retq ; ; SKX-LABEL: sext_32x8mem_to_32x16: @@ -213,7 +213,7 @@ define <32 x i16> @zext_32x8_to_32x16(<32 x i8> %a ) nounwind readnone { ; KNL-NEXT: vpmovzxbw {{.*#+}} ymm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero ; KNL-NEXT: vextracti128 $1, %ymm0, %xmm0 ; KNL-NEXT: vpmovzxbw {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero -; KNL-NEXT: vmovdqa64 %zmm2, %zmm0 +; KNL-NEXT: vmovdqa %ymm2, %ymm0 ; KNL-NEXT: retq ; ; SKX-LABEL: zext_32x8_to_32x16: @@ -258,7 +258,7 @@ define <32 x i16> @sext_32x8_to_32x16(<32 x i8> %a ) nounwind readnone { ; KNL-NEXT: vpmovsxbw %xmm0, %ymm2 ; KNL-NEXT: vextracti128 $1, %ymm0, %xmm0 ; KNL-NEXT: vpmovsxbw %xmm0, %ymm1 -; KNL-NEXT: vmovdqa64 %zmm2, %zmm0 +; KNL-NEXT: vmovdqa %ymm2, %ymm0 ; KNL-NEXT: retq ; ; SKX-LABEL: sext_32x8_to_32x16: |
