[llvm] r360517 - [X86][SSE] Add HADD/HSUB SimplifyDemandedVectorElts tests
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Sat May 11 05:46:38 PDT 2019
Author: rksimon
Date: Sat May 11 05:46:38 2019
New Revision: 360517
URL: http://llvm.org/viewvc/llvm-project?rev=360517&view=rev
Log:
[X86][SSE] Add HADD/HSUB SimplifyDemandedVectorElts tests
Shows missed opportunities to simplify args.
Will add integer HADD/HSUB tests in a future commit.
Added:
llvm/trunk/test/CodeGen/X86/horizontal-shuffle-demanded.ll
Added: llvm/trunk/test/CodeGen/X86/horizontal-shuffle-demanded.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/horizontal-shuffle-demanded.ll?rev=360517&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/horizontal-shuffle-demanded.ll (added)
+++ llvm/trunk/test/CodeGen/X86/horizontal-shuffle-demanded.ll Sat May 11 05:46:38 2019
@@ -0,0 +1,223 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=i686-apple-darwin -mattr=+avx2 | FileCheck %s --check-prefix=X86
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx2 | FileCheck %s --check-prefix=X64
+
+;
+; 128-bit Vectors
+;
+
+define void @test_demanded_haddps_128(<4 x float> %a0, <4 x float> %a1, float *%a2) nounwind {
+; X86-LABEL: test_demanded_haddps_128:
+; X86: ## %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: vbroadcastss %xmm1, %xmm1
+; X86-NEXT: vhaddps %xmm1, %xmm0, %xmm0
+; X86-NEXT: vmovss %xmm0, (%eax)
+; X86-NEXT: retl
+;
+; X64-LABEL: test_demanded_haddps_128:
+; X64: ## %bb.0:
+; X64-NEXT: vbroadcastss %xmm1, %xmm1
+; X64-NEXT: vhaddps %xmm1, %xmm0, %xmm0
+; X64-NEXT: vmovss %xmm0, (%rdi)
+; X64-NEXT: retq
+ %1 = shufflevector <4 x float> %a1, <4 x float> undef, <4 x i32> zeroinitializer
+ %2 = call <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float> %a0, <4 x float> %1)
+ %3 = extractelement <4 x float> %2, i32 0
+ store float %3, float *%a2
+ ret void
+}
+
+define void @test_demanded_hsubps_128(<4 x float> %a0, <4 x float> %a1, float *%a2) nounwind {
+; X86-LABEL: test_demanded_hsubps_128:
+; X86: ## %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: vbroadcastss %xmm1, %xmm1
+; X86-NEXT: vhsubps %xmm1, %xmm0, %xmm0
+; X86-NEXT: vextractps $1, %xmm0, (%eax)
+; X86-NEXT: retl
+;
+; X64-LABEL: test_demanded_hsubps_128:
+; X64: ## %bb.0:
+; X64-NEXT: vbroadcastss %xmm1, %xmm1
+; X64-NEXT: vhsubps %xmm1, %xmm0, %xmm0
+; X64-NEXT: vextractps $1, %xmm0, (%rdi)
+; X64-NEXT: retq
+ %1 = shufflevector <4 x float> %a1, <4 x float> undef, <4 x i32> zeroinitializer
+ %2 = call <4 x float> @llvm.x86.sse3.hsub.ps(<4 x float> %a0, <4 x float> %1)
+ %3 = extractelement <4 x float> %2, i32 1
+ store float %3, float *%a2
+ ret void
+}
+
+define void @test_demanded_haddpd_128(<2 x double> %a0, <2 x double> %a1, double *%a2) nounwind {
+; X86-LABEL: test_demanded_haddpd_128:
+; X86: ## %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: vmovddup {{.*#+}} xmm1 = xmm1[0,0]
+; X86-NEXT: vhaddpd %xmm1, %xmm0, %xmm0
+; X86-NEXT: vmovlpd %xmm0, (%eax)
+; X86-NEXT: retl
+;
+; X64-LABEL: test_demanded_haddpd_128:
+; X64: ## %bb.0:
+; X64-NEXT: vmovddup {{.*#+}} xmm1 = xmm1[0,0]
+; X64-NEXT: vhaddpd %xmm1, %xmm0, %xmm0
+; X64-NEXT: vmovlpd %xmm0, (%rdi)
+; X64-NEXT: retq
+ %1 = shufflevector <2 x double> %a1, <2 x double> undef, <2 x i32> zeroinitializer
+ %2 = call <2 x double> @llvm.x86.sse3.hadd.pd(<2 x double> %a0, <2 x double> %1)
+ %3 = extractelement <2 x double> %2, i32 0
+ store double %3, double *%a2
+ ret void
+}
+
+define void @test_demanded_hsubpd_128(<2 x double> %a0, <2 x double> %a1, double *%a2) nounwind {
+; X86-LABEL: test_demanded_hsubpd_128:
+; X86: ## %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: vmovddup {{.*#+}} xmm1 = xmm1[0,0]
+; X86-NEXT: vhsubpd %xmm1, %xmm0, %xmm0
+; X86-NEXT: vmovlpd %xmm0, (%eax)
+; X86-NEXT: retl
+;
+; X64-LABEL: test_demanded_hsubpd_128:
+; X64: ## %bb.0:
+; X64-NEXT: vmovddup {{.*#+}} xmm1 = xmm1[0,0]
+; X64-NEXT: vhsubpd %xmm1, %xmm0, %xmm0
+; X64-NEXT: vmovlpd %xmm0, (%rdi)
+; X64-NEXT: retq
+ %1 = shufflevector <2 x double> %a1, <2 x double> undef, <2 x i32> zeroinitializer
+ %2 = call <2 x double> @llvm.x86.sse3.hsub.pd(<2 x double> %a0, <2 x double> %1)
+ %3 = extractelement <2 x double> %2, i32 0
+ store double %3, double *%a2
+ ret void
+}
+
+;
+; 256-bit Vectors
+;
+
+define void @test_demanded_haddps_256(<8 x float> %a0, <8 x float> %a1, float *%a2) nounwind {
+; X86-LABEL: test_demanded_haddps_256:
+; X86: ## %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: vbroadcastss %xmm1, %ymm1
+; X86-NEXT: vhaddps %ymm1, %ymm0, %ymm0
+; X86-NEXT: vextractf128 $1, %ymm0, %xmm0
+; X86-NEXT: vmovss %xmm0, (%eax)
+; X86-NEXT: vzeroupper
+; X86-NEXT: retl
+;
+; X64-LABEL: test_demanded_haddps_256:
+; X64: ## %bb.0:
+; X64-NEXT: vbroadcastss %xmm1, %ymm1
+; X64-NEXT: vhaddps %ymm1, %ymm0, %ymm0
+; X64-NEXT: vextractf128 $1, %ymm0, %xmm0
+; X64-NEXT: vmovss %xmm0, (%rdi)
+; X64-NEXT: vzeroupper
+; X64-NEXT: retq
+ %1 = shufflevector <8 x float> %a1, <8 x float> undef, <8 x i32> zeroinitializer
+ %2 = call <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> %a0, <8 x float> %1)
+ %3 = extractelement <8 x float> %2, i32 4
+ store float %3, float *%a2
+ ret void
+}
+
+define void @test_demanded_hsubps_256(<8 x float> %a0, <8 x float> %a1, float *%a2) nounwind {
+; X86-LABEL: test_demanded_hsubps_256:
+; X86: ## %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: vbroadcastss %xmm1, %ymm1
+; X86-NEXT: vhsubps %ymm1, %ymm0, %ymm0
+; X86-NEXT: vextractf128 $1, %ymm0, %xmm0
+; X86-NEXT: vextractps $1, %xmm0, (%eax)
+; X86-NEXT: vzeroupper
+; X86-NEXT: retl
+;
+; X64-LABEL: test_demanded_hsubps_256:
+; X64: ## %bb.0:
+; X64-NEXT: vbroadcastss %xmm1, %ymm1
+; X64-NEXT: vhsubps %ymm1, %ymm0, %ymm0
+; X64-NEXT: vextractf128 $1, %ymm0, %xmm0
+; X64-NEXT: vextractps $1, %xmm0, (%rdi)
+; X64-NEXT: vzeroupper
+; X64-NEXT: retq
+ %1 = shufflevector <8 x float> %a1, <8 x float> undef, <8 x i32> zeroinitializer
+ %2 = call <8 x float> @llvm.x86.avx.hsub.ps.256(<8 x float> %a0, <8 x float> %1)
+ %3 = extractelement <8 x float> %2, i32 5
+ store float %3, float *%a2
+ ret void
+}
+
+define void @test_demanded_haddpd_256(<4 x double> %a0, <4 x double> %a1, double *%a2) nounwind {
+; X86-LABEL: test_demanded_haddpd_256:
+; X86: ## %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: vbroadcastsd %xmm1, %ymm1
+; X86-NEXT: vhaddpd %ymm1, %ymm0, %ymm0
+; X86-NEXT: vextractf128 $1, %ymm0, %xmm0
+; X86-NEXT: vmovlpd %xmm0, (%eax)
+; X86-NEXT: vzeroupper
+; X86-NEXT: retl
+;
+; X64-LABEL: test_demanded_haddpd_256:
+; X64: ## %bb.0:
+; X64-NEXT: vbroadcastsd %xmm1, %ymm1
+; X64-NEXT: vhaddpd %ymm1, %ymm0, %ymm0
+; X64-NEXT: vextractf128 $1, %ymm0, %xmm0
+; X64-NEXT: vmovlpd %xmm0, (%rdi)
+; X64-NEXT: vzeroupper
+; X64-NEXT: retq
+ %1 = shufflevector <4 x double> %a1, <4 x double> undef, <4 x i32> zeroinitializer
+ %2 = call <4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double> %a0, <4 x double> %1)
+ %3 = extractelement <4 x double> %2, i32 2
+ store double %3, double *%a2
+ ret void
+}
+
+define void @test_demanded_hsubpd_256(<4 x double> %a0, <4 x double> %a1, double *%a2) nounwind {
+; X86-LABEL: test_demanded_hsubpd_256:
+; X86: ## %bb.0:
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: vbroadcastsd %xmm1, %ymm1
+; X86-NEXT: vhsubpd %ymm1, %ymm0, %ymm0
+; X86-NEXT: vextractf128 $1, %ymm0, %xmm0
+; X86-NEXT: vmovlpd %xmm0, (%eax)
+; X86-NEXT: vzeroupper
+; X86-NEXT: retl
+;
+; X64-LABEL: test_demanded_hsubpd_256:
+; X64: ## %bb.0:
+; X64-NEXT: vbroadcastsd %xmm1, %ymm1
+; X64-NEXT: vhsubpd %ymm1, %ymm0, %ymm0
+; X64-NEXT: vextractf128 $1, %ymm0, %xmm0
+; X64-NEXT: vmovlpd %xmm0, (%rdi)
+; X64-NEXT: vzeroupper
+; X64-NEXT: retq
+ %1 = shufflevector <4 x double> %a1, <4 x double> undef, <4 x i32> zeroinitializer
+ %2 = call <4 x double> @llvm.x86.avx.hsub.pd.256(<4 x double> %a0, <4 x double> %1)
+ %3 = extractelement <4 x double> %2, i32 2
+ store double %3, double *%a2
+ ret void
+}
+
+declare <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float>, <4 x float>)
+declare <4 x float> @llvm.x86.sse3.hsub.ps(<4 x float>, <4 x float>)
+declare <2 x double> @llvm.x86.sse3.hadd.pd(<2 x double>, <2 x double>)
+declare <2 x double> @llvm.x86.sse3.hsub.pd(<2 x double>, <2 x double>)
+
+declare <8 x i16> @llvm.x86.ssse3.phadd.w.128(<8 x i16>, <8 x i16>)
+declare <4 x i32> @llvm.x86.ssse3.phadd.d.128(<4 x i32>, <4 x i32>)
+declare <8 x i16> @llvm.x86.ssse3.phsub.w.128(<8 x i16>, <8 x i16>)
+declare <4 x i32> @llvm.x86.ssse3.phsub.d.128(<4 x i32>, <4 x i32>)
+
+declare <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float>, <8 x float>)
+declare <8 x float> @llvm.x86.avx.hsub.ps.256(<8 x float>, <8 x float>)
+declare <4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double>, <4 x double>)
+declare <4 x double> @llvm.x86.avx.hsub.pd.256(<4 x double>, <4 x double>)
+
+declare <16 x i16> @llvm.x86.avx2.phadd.w(<16 x i16>, <16 x i16>)
+declare <8 x i32> @llvm.x86.avx2.phadd.d(<8 x i32>, <8 x i32>)
+declare <16 x i16> @llvm.x86.avx2.phsub.w(<16 x i16>, <16 x i16>)
+declare <8 x i32> @llvm.x86.avx2.phsub.d(<8 x i32>, <8 x i32>)
More information about the llvm-commits
mailing list