[llvm] r314626 - pre-commit adding test for broadcastm pattern
Jina Nahias via llvm-commits
llvm-commits at lists.llvm.org
Sun Oct 1 07:25:21 PDT 2017
Author: jina.nahias
Date: Sun Oct 1 07:25:21 2017
New Revision: 314626
URL: http://llvm.org/viewvc/llvm-project?rev=314626&view=rev
Log:
pre-commit adding test for broadcastm pattern
Differential Revision: https://reviews.llvm.org/D38312
Change-Id: Ifbc4189549f2f59995019a86f85f989c04e4d37d
Added:
llvm/trunk/test/CodeGen/X86/broadcastm-lowering.ll
Added: llvm/trunk/test/CodeGen/X86/broadcastm-lowering.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/broadcastm-lowering.ll?rev=314626&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/broadcastm-lowering.ll (added)
+++ llvm/trunk/test/CodeGen/X86/broadcastm-lowering.ll Sun Oct 1 07:25:21 2017
@@ -0,0 +1,263 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512cd | FileCheck %s --check-prefix=ALL --check-prefix=AVX512CD
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,avx512cd,+avx512bw| FileCheck %s --check-prefix=ALL --check-prefix=AVX512VLCDBW
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512vl,avx512cd,+avx512bw| FileCheck %s --check-prefix=ALL --check-prefix=X86-AVX512VLCDBW
+
+define <2 x i64> @test_mm_epi64(<8 x i16> %a, <8 x i16> %b) {
+; AVX512CD-LABEL: test_mm_epi64:
+; AVX512CD: # BB#0: # %entry
+; AVX512CD-NEXT: vpcmpeqw %xmm1, %xmm0, %xmm0
+; AVX512CD-NEXT: vpmovsxwq %xmm0, %zmm0
+; AVX512CD-NEXT: vpsllq $63, %zmm0, %zmm0
+; AVX512CD-NEXT: vptestmq %zmm0, %zmm0, %k0
+; AVX512CD-NEXT: kmovw %k0, %eax
+; AVX512CD-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX512CD-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0
+; AVX512CD-NEXT: vpinsrb $8, %eax, %xmm0, %xmm0
+; AVX512CD-NEXT: vzeroupper
+; AVX512CD-NEXT: retq
+;
+; AVX512VLCDBW-LABEL: test_mm_epi64:
+; AVX512VLCDBW: # BB#0: # %entry
+; AVX512VLCDBW-NEXT: vpcmpeqw %xmm1, %xmm0, %k0
+; AVX512VLCDBW-NEXT: kmovd %k0, %eax
+; AVX512VLCDBW-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrb $8, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: retq
+;
+; X86-AVX512VLCDBW-LABEL: test_mm_epi64:
+; X86-AVX512VLCDBW: # BB#0: # %entry
+; X86-AVX512VLCDBW-NEXT: vpcmpeqw %xmm1, %xmm0, %k0
+; X86-AVX512VLCDBW-NEXT: kmovd %k0, %eax
+; X86-AVX512VLCDBW-NEXT: movzbl %al, %eax
+; X86-AVX512VLCDBW-NEXT: vmovd %eax, %xmm0
+; X86-AVX512VLCDBW-NEXT: vpbroadcastq %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: retl
+entry:
+ %0 = icmp eq <8 x i16> %a, %b
+ %1 = bitcast <8 x i1> %0 to i8
+ %conv.i = zext i8 %1 to i64
+ %vecinit.i.i = insertelement <2 x i64> undef, i64 %conv.i, i32 0
+ %vecinit1.i.i = shufflevector <2 x i64> %vecinit.i.i, <2 x i64> undef, <2 x i32> zeroinitializer
+ ret <2 x i64> %vecinit1.i.i
+}
+
+define <4 x i32> @test_mm_epi32(<16 x i8> %a, <16 x i8> %b) {
+; AVX512CD-LABEL: test_mm_epi32:
+; AVX512CD: # BB#0: # %entry
+; AVX512CD-NEXT: vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX512CD-NEXT: vpmovsxbd %xmm0, %zmm0
+; AVX512CD-NEXT: vpslld $31, %zmm0, %zmm0
+; AVX512CD-NEXT: vptestmd %zmm0, %zmm0, %k0
+; AVX512CD-NEXT: kmovw %k0, %eax
+; AVX512CD-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX512CD-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; AVX512CD-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0
+; AVX512CD-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0
+; AVX512CD-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0
+; AVX512CD-NEXT: vzeroupper
+; AVX512CD-NEXT: retq
+;
+; AVX512VLCDBW-LABEL: test_mm_epi32:
+; AVX512VLCDBW: # BB#0: # %entry
+; AVX512VLCDBW-NEXT: vpcmpeqb %xmm1, %xmm0, %k0
+; AVX512VLCDBW-NEXT: kmovd %k0, %eax
+; AVX512VLCDBW-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: retq
+;
+; X86-AVX512VLCDBW-LABEL: test_mm_epi32:
+; X86-AVX512VLCDBW: # BB#0: # %entry
+; X86-AVX512VLCDBW-NEXT: vpcmpeqb %xmm1, %xmm0, %k0
+; X86-AVX512VLCDBW-NEXT: kmovd %k0, %eax
+; X86-AVX512VLCDBW-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: retl
+entry:
+ %0 = icmp eq <16 x i8> %a, %b
+ %1 = bitcast <16 x i1> %0 to i16
+ %conv.i = zext i16 %1 to i32
+ %vecinit.i.i = insertelement <4 x i32> undef, i32 %conv.i, i32 0
+ %vecinit3.i.i = shufflevector <4 x i32> %vecinit.i.i, <4 x i32> undef, <4 x i32> zeroinitializer
+ ret <4 x i32> %vecinit3.i.i
+}
+
+define <16 x i32> @test_mm512_epi32(<16 x i32> %a, <16 x i32> %b) {
+; AVX512CD-LABEL: test_mm512_epi32:
+; AVX512CD: # BB#0: # %entry
+; AVX512CD-NEXT: vpcmpeqd %zmm1, %zmm0, %k0
+; AVX512CD-NEXT: kmovw %k0, %eax
+; AVX512CD-NEXT: vpbroadcastd %eax, %zmm0
+; AVX512CD-NEXT: retq
+;
+; AVX512VLCDBW-LABEL: test_mm512_epi32:
+; AVX512VLCDBW: # BB#0: # %entry
+; AVX512VLCDBW-NEXT: vpcmpeqd %zmm1, %zmm0, %k0
+; AVX512VLCDBW-NEXT: kmovd %k0, %eax
+; AVX512VLCDBW-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512VLCDBW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
+; AVX512VLCDBW-NEXT: retq
+;
+; X86-AVX512VLCDBW-LABEL: test_mm512_epi32:
+; X86-AVX512VLCDBW: # BB#0: # %entry
+; X86-AVX512VLCDBW-NEXT: vpcmpeqd %zmm1, %zmm0, %k0
+; X86-AVX512VLCDBW-NEXT: kmovd %k0, %eax
+; X86-AVX512VLCDBW-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X86-AVX512VLCDBW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
+; X86-AVX512VLCDBW-NEXT: retl
+entry:
+ %0 = icmp eq <16 x i32> %a, %b
+ %1 = bitcast <16 x i1> %0 to i16
+ %conv.i = zext i16 %1 to i32
+ %vecinit.i.i = insertelement <16 x i32> undef, i32 %conv.i, i32 0
+ %vecinit15.i.i = shufflevector <16 x i32> %vecinit.i.i, <16 x i32> undef, <16 x i32> zeroinitializer
+ ret <16 x i32> %vecinit15.i.i
+}
+
+define <8 x i64> @test_mm512_epi64(<8 x i32> %a, <8 x i32> %b) {
+; AVX512CD-LABEL: test_mm512_epi64:
+; AVX512CD: # BB#0: # %entry
+; AVX512CD-NEXT: # kill: %YMM1<def> %YMM1<kill> %ZMM1<def>
+; AVX512CD-NEXT: # kill: %YMM0<def> %YMM0<kill> %ZMM0<def>
+; AVX512CD-NEXT: vpcmpeqd %zmm1, %zmm0, %k0
+; AVX512CD-NEXT: kmovw %k0, %eax
+; AVX512CD-NEXT: movzbl %al, %eax
+; AVX512CD-NEXT: vpbroadcastq %rax, %zmm0
+; AVX512CD-NEXT: retq
+;
+; AVX512VLCDBW-LABEL: test_mm512_epi64:
+; AVX512VLCDBW: # BB#0: # %entry
+; AVX512VLCDBW-NEXT: vpcmpeqd %ymm1, %ymm0, %k0
+; AVX512VLCDBW-NEXT: kmovd %k0, %eax
+; AVX512VLCDBW-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrb $8, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512VLCDBW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
+; AVX512VLCDBW-NEXT: retq
+;
+; X86-AVX512VLCDBW-LABEL: test_mm512_epi64:
+; X86-AVX512VLCDBW: # BB#0: # %entry
+; X86-AVX512VLCDBW-NEXT: vpcmpeqd %ymm1, %ymm0, %k0
+; X86-AVX512VLCDBW-NEXT: kmovd %k0, %eax
+; X86-AVX512VLCDBW-NEXT: movzbl %al, %eax
+; X86-AVX512VLCDBW-NEXT: vmovd %eax, %xmm0
+; X86-AVX512VLCDBW-NEXT: vpbroadcastq %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X86-AVX512VLCDBW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
+; X86-AVX512VLCDBW-NEXT: retl
+entry:
+ %0 = icmp eq <8 x i32> %a, %b
+ %1 = bitcast <8 x i1> %0 to i8
+ %conv.i = zext i8 %1 to i64
+ %vecinit.i.i = insertelement <8 x i64> undef, i64 %conv.i, i32 0
+ %vecinit7.i.i = shufflevector <8 x i64> %vecinit.i.i, <8 x i64> undef, <8 x i32> zeroinitializer
+ ret <8 x i64> %vecinit7.i.i
+}
+
+define <4 x i64> @test_mm256_epi64(<8 x i32> %a, <8 x i32> %b) {
+; AVX512CD-LABEL: test_mm256_epi64:
+; AVX512CD: # BB#0: # %entry
+; AVX512CD-NEXT: # kill: %YMM1<def> %YMM1<kill> %ZMM1<def>
+; AVX512CD-NEXT: # kill: %YMM0<def> %YMM0<kill> %ZMM0<def>
+; AVX512CD-NEXT: vpcmpeqd %zmm1, %zmm0, %k0
+; AVX512CD-NEXT: kmovw %k0, %eax
+; AVX512CD-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX512CD-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0
+; AVX512CD-NEXT: vpinsrb $8, %eax, %xmm0, %xmm0
+; AVX512CD-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512CD-NEXT: retq
+;
+; AVX512VLCDBW-LABEL: test_mm256_epi64:
+; AVX512VLCDBW: # BB#0: # %entry
+; AVX512VLCDBW-NEXT: vpcmpeqd %ymm1, %ymm0, %k0
+; AVX512VLCDBW-NEXT: kmovd %k0, %eax
+; AVX512VLCDBW-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrb $8, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512VLCDBW-NEXT: retq
+;
+; X86-AVX512VLCDBW-LABEL: test_mm256_epi64:
+; X86-AVX512VLCDBW: # BB#0: # %entry
+; X86-AVX512VLCDBW-NEXT: vpcmpeqd %ymm1, %ymm0, %k0
+; X86-AVX512VLCDBW-NEXT: kmovd %k0, %eax
+; X86-AVX512VLCDBW-NEXT: movzbl %al, %eax
+; X86-AVX512VLCDBW-NEXT: vmovd %eax, %xmm0
+; X86-AVX512VLCDBW-NEXT: vpbroadcastq %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X86-AVX512VLCDBW-NEXT: retl
+entry:
+ %0 = icmp eq <8 x i32> %a, %b
+ %1 = bitcast <8 x i1> %0 to i8
+ %conv.i = zext i8 %1 to i64
+ %vecinit.i.i = insertelement <4 x i64> undef, i64 %conv.i, i32 0
+ %vecinit3.i.i = shufflevector <4 x i64> %vecinit.i.i, <4 x i64> undef, <4 x i32> zeroinitializer
+ ret <4 x i64> %vecinit3.i.i
+}
+
+define <8 x i32> @test_mm256_epi32(<16 x i16> %a, <16 x i16> %b) {
+; AVX512CD-LABEL: test_mm256_epi32:
+; AVX512CD: # BB#0: # %entry
+; AVX512CD-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0
+; AVX512CD-NEXT: vpmovsxwd %ymm0, %zmm0
+; AVX512CD-NEXT: vpslld $31, %zmm0, %zmm0
+; AVX512CD-NEXT: vptestmd %zmm0, %zmm0, %k0
+; AVX512CD-NEXT: kmovw %k0, %eax
+; AVX512CD-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX512CD-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; AVX512CD-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0
+; AVX512CD-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0
+; AVX512CD-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0
+; AVX512CD-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512CD-NEXT: retq
+;
+; AVX512VLCDBW-LABEL: test_mm256_epi32:
+; AVX512VLCDBW: # BB#0: # %entry
+; AVX512VLCDBW-NEXT: vpcmpeqw %ymm1, %ymm0, %k0
+; AVX512VLCDBW-NEXT: kmovd %k0, %eax
+; AVX512VLCDBW-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0
+; AVX512VLCDBW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX512VLCDBW-NEXT: retq
+;
+; X86-AVX512VLCDBW-LABEL: test_mm256_epi32:
+; X86-AVX512VLCDBW: # BB#0: # %entry
+; X86-AVX512VLCDBW-NEXT: vpcmpeqw %ymm1, %ymm0, %k0
+; X86-AVX512VLCDBW-NEXT: kmovd %k0, %eax
+; X86-AVX512VLCDBW-NEXT: vpxor %xmm0, %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0
+; X86-AVX512VLCDBW-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
+; X86-AVX512VLCDBW-NEXT: retl
+entry:
+ %0 = icmp eq <16 x i16> %a, %b
+ %1 = bitcast <16 x i1> %0 to i16
+ %conv.i = zext i16 %1 to i32
+ %vecinit.i.i = insertelement <8 x i32> undef, i32 %conv.i, i32 0
+ %vecinit7.i.i = shufflevector <8 x i32> %vecinit.i.i, <8 x i32> undef, <8 x i32> zeroinitializer
+ ret <8 x i32> %vecinit7.i.i
+}
+
More information about the llvm-commits
mailing list