diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/shuffle-vs-trunc-128.ll')
-rw-r--r-- | llvm/test/CodeGen/X86/shuffle-vs-trunc-128.ll | 159 |
1 files changed, 33 insertions, 126 deletions
diff --git a/llvm/test/CodeGen/X86/shuffle-vs-trunc-128.ll b/llvm/test/CodeGen/X86/shuffle-vs-trunc-128.ll index 9c5344562a9..7805c46e5c3 100644 --- a/llvm/test/CodeGen/X86/shuffle-vs-trunc-128.ll +++ b/llvm/test/CodeGen/X86/shuffle-vs-trunc-128.ll @@ -37,32 +37,12 @@ define void @shuffle_v16i8_to_v8i8(<16 x i8>* %L, <8 x i8>* %S) nounwind { ; AVX-NEXT: vmovq %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512F-LABEL: shuffle_v16i8_to_v8i8: -; AVX512F: # %bb.0: -; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u] -; AVX512F-NEXT: vmovq %xmm0, (%rsi) -; AVX512F-NEXT: retq -; -; AVX512VL-LABEL: shuffle_v16i8_to_v8i8: -; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512VL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u] -; AVX512VL-NEXT: vmovq %xmm0, (%rsi) -; AVX512VL-NEXT: retq -; -; AVX512BW-LABEL: shuffle_v16i8_to_v8i8: -; AVX512BW: # %bb.0: -; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u] -; AVX512BW-NEXT: vmovq %xmm0, (%rsi) -; AVX512BW-NEXT: retq -; -; AVX512BWVL-LABEL: shuffle_v16i8_to_v8i8: -; AVX512BWVL: # %bb.0: -; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512BWVL-NEXT: vpmovwb %xmm0, (%rsi) -; AVX512BWVL-NEXT: retq +; AVX512-LABEL: shuffle_v16i8_to_v8i8: +; AVX512: # %bb.0: +; AVX512-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u] +; AVX512-NEXT: vmovq %xmm0, (%rsi) +; AVX512-NEXT: retq %vec = load <16 x i8>, <16 x i8>* %L %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> store <8 x i8> %strided.vec, <8 x i8>* %S @@ -148,31 +128,12 @@ define void @shuffle_v8i16_to_v4i16(<8 x i16>* %L, <4 x i16>* %S) nounwind { ; AVX-NEXT: vmovq %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512F-LABEL: shuffle_v8i16_to_v4i16: -; AVX512F: # %bb.0: -; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15] -; AVX512F-NEXT: vmovq %xmm0, (%rsi) -; AVX512F-NEXT: retq -; -; AVX512VL-LABEL: shuffle_v8i16_to_v4i16: -; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512VL-NEXT: vpmovdw %xmm0, (%rsi) -; AVX512VL-NEXT: retq -; -; AVX512BW-LABEL: shuffle_v8i16_to_v4i16: -; AVX512BW: # %bb.0: -; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15] -; AVX512BW-NEXT: vmovq %xmm0, (%rsi) -; AVX512BW-NEXT: retq -; -; AVX512BWVL-LABEL: shuffle_v8i16_to_v4i16: -; AVX512BWVL: # %bb.0: -; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512BWVL-NEXT: vpmovdw %xmm0, (%rsi) -; AVX512BWVL-NEXT: retq +; AVX512-LABEL: shuffle_v8i16_to_v4i16: +; AVX512: # %bb.0: +; AVX512-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15] +; AVX512-NEXT: vmovq %xmm0, (%rsi) +; AVX512-NEXT: retq %vec = load <8 x i16>, <8 x i16>* %L %strided.vec = shufflevector <8 x i16> %vec, <8 x i16> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6> store <4 x i16> %strided.vec, <4 x i16>* %S @@ -247,29 +208,11 @@ define void @shuffle_v4i32_to_v2i32(<4 x i32>* %L, <2 x i32>* %S) nounwind { ; AVX-NEXT: vmovlps %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512F-LABEL: shuffle_v4i32_to_v2i32: -; AVX512F: # %bb.0: -; AVX512F-NEXT: vpermilps {{.*#+}} xmm0 = mem[0,2,2,3] -; AVX512F-NEXT: vmovlps %xmm0, (%rsi) -; AVX512F-NEXT: retq -; -; AVX512VL-LABEL: shuffle_v4i32_to_v2i32: -; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512VL-NEXT: vpmovqd %xmm0, (%rsi) -; AVX512VL-NEXT: retq -; -; AVX512BW-LABEL: shuffle_v4i32_to_v2i32: -; AVX512BW: # %bb.0: -; AVX512BW-NEXT: vpermilps {{.*#+}} xmm0 = mem[0,2,2,3] -; AVX512BW-NEXT: vmovlps %xmm0, (%rsi) -; AVX512BW-NEXT: retq -; -; AVX512BWVL-LABEL: shuffle_v4i32_to_v2i32: -; AVX512BWVL: # %bb.0: -; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512BWVL-NEXT: vpmovqd %xmm0, (%rsi) -; AVX512BWVL-NEXT: retq +; AVX512-LABEL: shuffle_v4i32_to_v2i32: +; AVX512: # %bb.0: +; AVX512-NEXT: vpermilps {{.*#+}} xmm0 = mem[0,2,2,3] +; AVX512-NEXT: vmovlps %xmm0, (%rsi) +; AVX512-NEXT: retq %vec = load <4 x i32>, <4 x i32>* %L %strided.vec = shufflevector <4 x i32> %vec, <4 x i32> undef, <2 x i32> <i32 0, i32 2> store <2 x i32> %strided.vec, <2 x i32>* %S @@ -343,31 +286,12 @@ define void @shuffle_v16i8_to_v4i8(<16 x i8>* %L, <4 x i8>* %S) nounwind { ; AVX-NEXT: vmovd %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512F-LABEL: shuffle_v16i8_to_v4i8: -; AVX512F: # %bb.0: -; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u] -; AVX512F-NEXT: vmovd %xmm0, (%rsi) -; AVX512F-NEXT: retq -; -; AVX512VL-LABEL: shuffle_v16i8_to_v4i8: -; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512VL-NEXT: vpmovdb %xmm0, (%rsi) -; AVX512VL-NEXT: retq -; -; AVX512BW-LABEL: shuffle_v16i8_to_v4i8: -; AVX512BW: # %bb.0: -; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u] -; AVX512BW-NEXT: vmovd %xmm0, (%rsi) -; AVX512BW-NEXT: retq -; -; AVX512BWVL-LABEL: shuffle_v16i8_to_v4i8: -; AVX512BWVL: # %bb.0: -; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512BWVL-NEXT: vpmovdb %xmm0, (%rsi) -; AVX512BWVL-NEXT: retq +; AVX512-LABEL: shuffle_v16i8_to_v4i8: +; AVX512: # %bb.0: +; AVX512-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512-NEXT: vmovd %xmm0, (%rsi) +; AVX512-NEXT: retq %vec = load <16 x i8>, <16 x i8>* %L %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12> store <4 x i8> %strided.vec, <4 x i8>* %S @@ -469,7 +393,8 @@ define void @shuffle_v8i16_to_v2i16(<8 x i16>* %L, <2 x i16>* %S) nounwind { ; AVX512VL-LABEL: shuffle_v8i16_to_v2i16: ; AVX512VL: # %bb.0: ; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512VL-NEXT: vpmovqw %xmm0, (%rsi) +; AVX512VL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,8,9,8,9,10,11,8,9,10,11,12,13,14,15] +; AVX512VL-NEXT: vmovd %xmm0, (%rsi) ; AVX512VL-NEXT: retq ; ; AVX512BW-LABEL: shuffle_v8i16_to_v2i16: @@ -482,7 +407,8 @@ define void @shuffle_v8i16_to_v2i16(<8 x i16>* %L, <2 x i16>* %S) nounwind { ; AVX512BWVL-LABEL: shuffle_v8i16_to_v2i16: ; AVX512BWVL: # %bb.0: ; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512BWVL-NEXT: vpmovqw %xmm0, (%rsi) +; AVX512BWVL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,8,9,8,9,10,11,8,9,10,11,12,13,14,15] +; AVX512BWVL-NEXT: vmovd %xmm0, (%rsi) ; AVX512BWVL-NEXT: retq %vec = load <8 x i16>, <8 x i16>* %L %strided.vec = shufflevector <8 x i16> %vec, <8 x i16> undef, <2 x i32> <i32 0, i32 4> @@ -577,31 +503,12 @@ define void @shuffle_v16i8_to_v2i8(<16 x i8>* %L, <2 x i8>* %S) nounwind { ; AVX-NEXT: vpextrw $0, %xmm0, (%rsi) ; AVX-NEXT: retq ; -; AVX512F-LABEL: shuffle_v16i8_to_v2i8: -; AVX512F: # %bb.0: -; AVX512F-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512F-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,8,u,u,u,u,u,u,u,u,u,u,u,u,u,u] -; AVX512F-NEXT: vpextrw $0, %xmm0, (%rsi) -; AVX512F-NEXT: retq -; -; AVX512VL-LABEL: shuffle_v16i8_to_v2i8: -; AVX512VL: # %bb.0: -; AVX512VL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512VL-NEXT: vpmovqb %xmm0, (%rsi) -; AVX512VL-NEXT: retq -; -; AVX512BW-LABEL: shuffle_v16i8_to_v2i8: -; AVX512BW: # %bb.0: -; AVX512BW-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512BW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,8,u,u,u,u,u,u,u,u,u,u,u,u,u,u] -; AVX512BW-NEXT: vpextrw $0, %xmm0, (%rsi) -; AVX512BW-NEXT: retq -; -; AVX512BWVL-LABEL: shuffle_v16i8_to_v2i8: -; AVX512BWVL: # %bb.0: -; AVX512BWVL-NEXT: vmovdqa (%rdi), %xmm0 -; AVX512BWVL-NEXT: vpmovqb %xmm0, (%rsi) -; AVX512BWVL-NEXT: retq +; AVX512-LABEL: shuffle_v16i8_to_v2i8: +; AVX512: # %bb.0: +; AVX512-NEXT: vmovdqa (%rdi), %xmm0 +; AVX512-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,8,u,u,u,u,u,u,u,u,u,u,u,u,u,u] +; AVX512-NEXT: vpextrw $0, %xmm0, (%rsi) +; AVX512-NEXT: retq %vec = load <16 x i8>, <16 x i8>* %L %strided.vec = shufflevector <16 x i8> %vec, <16 x i8> undef, <2 x i32> <i32 0, i32 8> store <2 x i8> %strided.vec, <2 x i8>* %S |