From 69f2299efc5249affe4aaad0a50ba022298db73d Mon Sep 17 00:00:00 2001 From: Simon Pilgrim Date: Sun, 7 Aug 2016 12:41:36 +0000 Subject: [PATCH] [X86][AVX512BW] Add sext/zext AVX512BW 512-bit vector tests llvm-svn: 277957 --- llvm/test/CodeGen/X86/vector-sext.ll | 148 +++++++++++++++++++++++------------ llvm/test/CodeGen/X86/vector-zext.ll | 40 +++++++--- 2 files changed, 128 insertions(+), 60 deletions(-) diff --git a/llvm/test/CodeGen/X86/vector-sext.ll b/llvm/test/CodeGen/X86/vector-sext.ll index 4ff7500..a7138bc 100644 --- a/llvm/test/CodeGen/X86/vector-sext.ll +++ b/llvm/test/CodeGen/X86/vector-sext.ll @@ -5,6 +5,7 @@ ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX --check-prefix=AVX1 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX --check-prefix=AVX2 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefix=AVX --check-prefix=AVX512 --check-prefix=AVX512F +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefix=AVX --check-prefix=AVX512 --check-prefix=AVX512BW ; ; Just one 32-bit run to make sure we do reasonable things there. ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=X32-SSE41 @@ -162,13 +163,18 @@ define <32 x i16> @sext_32i8_to_32i16(<32 x i8> %A) nounwind uwtable readnone ss ; AVX2-NEXT: vmovdqa %ymm2, %ymm0 ; AVX2-NEXT: retq ; -; AVX512-LABEL: sext_32i8_to_32i16: -; AVX512: # BB#0: # %entry -; AVX512-NEXT: vpmovsxbw %xmm0, %ymm2 -; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX512-NEXT: vpmovsxbw %xmm0, %ymm1 -; AVX512-NEXT: vmovdqa64 %zmm2, %zmm0 -; AVX512-NEXT: retq +; AVX512F-LABEL: sext_32i8_to_32i16: +; AVX512F: # BB#0: # %entry +; AVX512F-NEXT: vpmovsxbw %xmm0, %ymm2 +; AVX512F-NEXT: vextracti128 $1, %ymm0, %xmm0 +; AVX512F-NEXT: vpmovsxbw %xmm0, %ymm1 +; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm0 +; AVX512F-NEXT: retq +; +; AVX512BW-LABEL: sext_32i8_to_32i16: +; AVX512BW: # BB#0: # %entry +; AVX512BW-NEXT: vpmovsxbw %ymm0, %zmm0 +; AVX512BW-NEXT: retq ; ; X32-SSE41-LABEL: sext_32i8_to_32i16: ; X32-SSE41: # BB#0: # %entry @@ -1241,14 +1247,23 @@ define <2 x i64> @load_sext_2i1_to_2i64(<2 x i1> *%ptr) { ; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0] ; AVX2-NEXT: retq ; -; AVX512-LABEL: load_sext_2i1_to_2i64: -; AVX512: # BB#0: # %entry -; AVX512-NEXT: movzbl (%rdi), %eax -; AVX512-NEXT: kmovw %eax, %k1 -; AVX512-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 -; AVX512-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z} -; AVX512-NEXT: # kill: %XMM0 %XMM0 %ZMM0 -; AVX512-NEXT: retq +; AVX512F-LABEL: load_sext_2i1_to_2i64: +; AVX512F: # BB#0: # %entry +; AVX512F-NEXT: movzbl (%rdi), %eax +; AVX512F-NEXT: kmovw %eax, %k1 +; AVX512F-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 +; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z} +; AVX512F-NEXT: # kill: %XMM0 %XMM0 %ZMM0 +; AVX512F-NEXT: retq +; +; AVX512BW-LABEL: load_sext_2i1_to_2i64: +; AVX512BW: # BB#0: # %entry +; AVX512BW-NEXT: movzbl (%rdi), %eax +; AVX512BW-NEXT: kmovd %eax, %k1 +; AVX512BW-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 +; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z} +; AVX512BW-NEXT: # kill: %XMM0 %XMM0 %ZMM0 +; AVX512BW-NEXT: retq ; ; X32-SSE41-LABEL: load_sext_2i1_to_2i64: ; X32-SSE41: # BB#0: # %entry @@ -1423,15 +1438,25 @@ define <4 x i32> @load_sext_4i1_to_4i32(<4 x i1> *%ptr) { ; AVX2-NEXT: vpinsrd $3, %eax, %xmm0, %xmm0 ; AVX2-NEXT: retq ; -; AVX512-LABEL: load_sext_4i1_to_4i32: -; AVX512: # BB#0: # %entry -; AVX512-NEXT: movzbl (%rdi), %eax -; AVX512-NEXT: kmovw %eax, %k1 -; AVX512-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 -; AVX512-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z} -; AVX512-NEXT: vpmovqd %zmm0, %ymm0 -; AVX512-NEXT: # kill: %XMM0 %XMM0 %YMM0 -; AVX512-NEXT: retq +; AVX512F-LABEL: load_sext_4i1_to_4i32: +; AVX512F: # BB#0: # %entry +; AVX512F-NEXT: movzbl (%rdi), %eax +; AVX512F-NEXT: kmovw %eax, %k1 +; AVX512F-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 +; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z} +; AVX512F-NEXT: vpmovqd %zmm0, %ymm0 +; AVX512F-NEXT: # kill: %XMM0 %XMM0 %YMM0 +; AVX512F-NEXT: retq +; +; AVX512BW-LABEL: load_sext_4i1_to_4i32: +; AVX512BW: # BB#0: # %entry +; AVX512BW-NEXT: movzbl (%rdi), %eax +; AVX512BW-NEXT: kmovd %eax, %k1 +; AVX512BW-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 +; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z} +; AVX512BW-NEXT: vpmovqd %zmm0, %ymm0 +; AVX512BW-NEXT: # kill: %XMM0 %XMM0 %YMM0 +; AVX512BW-NEXT: retq ; ; X32-SSE41-LABEL: load_sext_4i1_to_4i32: ; X32-SSE41: # BB#0: # %entry @@ -1620,14 +1645,23 @@ define <4 x i64> @load_sext_4i1_to_4i64(<4 x i1> *%ptr) { ; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 ; AVX2-NEXT: retq ; -; AVX512-LABEL: load_sext_4i1_to_4i64: -; AVX512: # BB#0: # %entry -; AVX512-NEXT: movzbl (%rdi), %eax -; AVX512-NEXT: kmovw %eax, %k1 -; AVX512-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 -; AVX512-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z} -; AVX512-NEXT: # kill: %YMM0 %YMM0 %ZMM0 -; AVX512-NEXT: retq +; AVX512F-LABEL: load_sext_4i1_to_4i64: +; AVX512F: # BB#0: # %entry +; AVX512F-NEXT: movzbl (%rdi), %eax +; AVX512F-NEXT: kmovw %eax, %k1 +; AVX512F-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 +; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z} +; AVX512F-NEXT: # kill: %YMM0 %YMM0 %ZMM0 +; AVX512F-NEXT: retq +; +; AVX512BW-LABEL: load_sext_4i1_to_4i64: +; AVX512BW: # BB#0: # %entry +; AVX512BW-NEXT: movzbl (%rdi), %eax +; AVX512BW-NEXT: kmovd %eax, %k1 +; AVX512BW-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 +; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z} +; AVX512BW-NEXT: # kill: %YMM0 %YMM0 %ZMM0 +; AVX512BW-NEXT: retq ; ; X32-SSE41-LABEL: load_sext_4i1_to_4i64: ; X32-SSE41: # BB#0: # %entry @@ -1914,14 +1948,23 @@ define <8 x i16> @load_sext_8i1_to_8i16(<8 x i1> *%ptr) { ; AVX2-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0 ; AVX2-NEXT: retq ; -; AVX512-LABEL: load_sext_8i1_to_8i16: -; AVX512: # BB#0: # %entry -; AVX512-NEXT: movzbl (%rdi), %eax -; AVX512-NEXT: kmovw %eax, %k1 -; AVX512-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 -; AVX512-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z} -; AVX512-NEXT: vpmovqw %zmm0, %xmm0 -; AVX512-NEXT: retq +; AVX512F-LABEL: load_sext_8i1_to_8i16: +; AVX512F: # BB#0: # %entry +; AVX512F-NEXT: movzbl (%rdi), %eax +; AVX512F-NEXT: kmovw %eax, %k1 +; AVX512F-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 +; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z} +; AVX512F-NEXT: vpmovqw %zmm0, %xmm0 +; AVX512F-NEXT: retq +; +; AVX512BW-LABEL: load_sext_8i1_to_8i16: +; AVX512BW: # BB#0: # %entry +; AVX512BW-NEXT: movzbl (%rdi), %eax +; AVX512BW-NEXT: kmovd %eax, %k1 +; AVX512BW-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 +; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z} +; AVX512BW-NEXT: vpmovqw %zmm0, %xmm0 +; AVX512BW-NEXT: retq ; ; X32-SSE41-LABEL: load_sext_8i1_to_8i16: ; X32-SSE41: # BB#0: # %entry @@ -2308,14 +2351,23 @@ define <8 x i32> @load_sext_8i1_to_8i32(<8 x i1> *%ptr) { ; AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 ; AVX2-NEXT: retq ; -; AVX512-LABEL: load_sext_8i1_to_8i32: -; AVX512: # BB#0: # %entry -; AVX512-NEXT: movzbl (%rdi), %eax -; AVX512-NEXT: kmovw %eax, %k1 -; AVX512-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 -; AVX512-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z} -; AVX512-NEXT: vpmovqd %zmm0, %ymm0 -; AVX512-NEXT: retq +; AVX512F-LABEL: load_sext_8i1_to_8i32: +; AVX512F: # BB#0: # %entry +; AVX512F-NEXT: movzbl (%rdi), %eax +; AVX512F-NEXT: kmovw %eax, %k1 +; AVX512F-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 +; AVX512F-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z} +; AVX512F-NEXT: vpmovqd %zmm0, %ymm0 +; AVX512F-NEXT: retq +; +; AVX512BW-LABEL: load_sext_8i1_to_8i32: +; AVX512BW: # BB#0: # %entry +; AVX512BW-NEXT: movzbl (%rdi), %eax +; AVX512BW-NEXT: kmovd %eax, %k1 +; AVX512BW-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 +; AVX512BW-NEXT: vmovdqa64 %zmm0, %zmm0 {%k1} {z} +; AVX512BW-NEXT: vpmovqd %zmm0, %ymm0 +; AVX512BW-NEXT: retq ; ; X32-SSE41-LABEL: load_sext_8i1_to_8i32: ; X32-SSE41: # BB#0: # %entry diff --git a/llvm/test/CodeGen/X86/vector-zext.ll b/llvm/test/CodeGen/X86/vector-zext.ll index f79b031..cb3583c 100644 --- a/llvm/test/CodeGen/X86/vector-zext.ll +++ b/llvm/test/CodeGen/X86/vector-zext.ll @@ -5,6 +5,7 @@ ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX --check-prefix=AVX1 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX --check-prefix=AVX2 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefix=AVX --check-prefix=AVX512 --check-prefix=AVX512F +; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefix=AVX --check-prefix=AVX512 --check-prefix=AVX512BW define <8 x i16> @zext_16i8_to_8i16(<16 x i8> %A) nounwind uwtable readnone ssp { ; SSE2-LABEL: zext_16i8_to_8i16: @@ -139,13 +140,18 @@ define <32 x i16> @zext_32i8_to_32i16(<32 x i8> %A) { ; AVX2-NEXT: vmovdqa %ymm2, %ymm0 ; AVX2-NEXT: retq ; -; AVX512-LABEL: zext_32i8_to_32i16: -; AVX512: # BB#0: # %entry -; AVX512-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 -; AVX512-NEXT: vextracti128 $1, %ymm0, %xmm0 -; AVX512-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 -; AVX512-NEXT: vmovdqa64 %zmm2, %zmm0 -; AVX512-NEXT: retq +; AVX512F-LABEL: zext_32i8_to_32i16: +; AVX512F: # BB#0: # %entry +; AVX512F-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 +; AVX512F-NEXT: vextracti128 $1, %ymm0, %xmm0 +; AVX512F-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 +; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm0 +; AVX512F-NEXT: retq +; +; AVX512BW-LABEL: zext_32i8_to_32i16: +; AVX512BW: # BB#0: # %entry +; AVX512BW-NEXT: vpmovzxbw {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero,ymm0[16],zero,ymm0[17],zero,ymm0[18],zero,ymm0[19],zero,ymm0[20],zero,ymm0[21],zero,ymm0[22],zero,ymm0[23],zero,ymm0[24],zero,ymm0[25],zero,ymm0[26],zero,ymm0[27],zero,ymm0[28],zero,ymm0[29],zero,ymm0[30],zero,ymm0[31],zero +; AVX512BW-NEXT: retq entry: %B = zext <32 x i8> %A to <32 x i16> ret <32 x i16> %B @@ -452,11 +458,21 @@ define <8 x i64> @zext_16i8_to_8i64(<16 x i8> %A) nounwind uwtable readnone ssp ; AVX2-NEXT: vmovdqa %ymm2, %ymm0 ; AVX2-NEXT: retq ; -; AVX512-LABEL: zext_16i8_to_8i64: -; AVX512: # BB#0: # %entry -; AVX512-NEXT: vpmovzxbq {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero,xmm0[4],zero,zero,zero,zero,zero,zero,zero,xmm0[5],zero,zero,zero,zero,zero,zero,zero,xmm0[6],zero,zero,zero,zero,zero,zero,zero,xmm0[7],zero,zero,zero,zero,zero,zero,zero -; AVX512-NEXT: vpandq {{.*}}(%rip){1to8}, %zmm0, %zmm0 -; AVX512-NEXT: retq +; AVX512F-LABEL: zext_16i8_to_8i64: +; AVX512F: # BB#0: # %entry +; AVX512F-NEXT: vpmovzxbq {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero,xmm0[4],zero,zero,zero,zero,zero,zero,zero,xmm0[5],zero,zero,zero,zero,zero,zero,zero,xmm0[6],zero,zero,zero,zero,zero,zero,zero,xmm0[7],zero,zero,zero,zero,zero,zero,zero +; AVX512F-NEXT: vpandq {{.*}}(%rip){1to8}, %zmm0, %zmm0 +; AVX512F-NEXT: retq +; +; AVX512BW-LABEL: zext_16i8_to_8i64: +; AVX512BW: # BB#0: # %entry +; AVX512BW-NEXT: vpmovzxbq {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero,xmm0[4],zero,zero,zero,zero,zero,zero,zero,xmm0[5],zero,zero,zero,zero,zero,zero,zero,xmm0[6],zero,zero,zero,zero,zero,zero,zero,xmm0[7],zero,zero,zero,zero,zero,zero,zero +; AVX512BW-NEXT: vextracti64x4 $1, %zmm0, %ymm1 +; AVX512BW-NEXT: vmovdqa {{.*#+}} ymm2 = [255,0,0,0,0,0,0,0,255,0,0,0,0,0,0,0,255,0,0,0,0,0,0,0,255,0,0,0,0,0,0,0] +; AVX512BW-NEXT: vpand %ymm2, %ymm1, %ymm1 +; AVX512BW-NEXT: vpand %ymm2, %ymm0, %ymm0 +; AVX512BW-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0 +; AVX512BW-NEXT: retq entry: %B = shufflevector <16 x i8> %A, <16 x i8> undef, <8 x i32> %C = zext <8 x i8> %B to <8 x i64> -- 2.7.4