[llvm] r345468 - [X86] Add test cases showing missed opportunities for optimizing vector fcopysign when the RHS is a splat constant.
Craig Topper via llvm-commits
llvm-commits at lists.llvm.org
Sat Oct 27 18:32:48 PDT 2018
Author: ctopper
Date: Sat Oct 27 18:32:47 2018
New Revision: 345468
URL: http://llvm.org/viewvc/llvm-project?rev=345468&view=rev
Log:
[X86] Add test cases showing missed opportunities for optimizing vector fcopysign when the RHS is a splat constant.
Added:
llvm/trunk/test/CodeGen/X86/sse1-fcopysign.ll
Added: llvm/trunk/test/CodeGen/X86/sse1-fcopysign.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/sse1-fcopysign.ll?rev=345468&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/sse1-fcopysign.ll (added)
+++ llvm/trunk/test/CodeGen/X86/sse1-fcopysign.ll Sat Oct 27 18:32:47 2018
@@ -0,0 +1,84 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=i686-unknown -mattr=-sse2,+sse | FileCheck %s --check-prefix=ALL --check-prefix=X86
+; RUN: llc < %s -mtriple=x86_64-unknown -mattr=-sse2,+sse | FileCheck %s --check-prefix=ALL --check-prefix=X64
+
+define float @f32_pos(float %a, float %b) nounwind {
+; X86-LABEL: f32_pos:
+; X86: # %bb.0:
+; X86-NEXT: pushl %eax
+; X86-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-NEXT: andps {{\.LCPI.*}}, %xmm0
+; X86-NEXT: movss %xmm0, (%esp)
+; X86-NEXT: flds (%esp)
+; X86-NEXT: popl %eax
+; X86-NEXT: retl
+;
+; X64-LABEL: f32_pos:
+; X64: # %bb.0:
+; X64-NEXT: andps {{.*}}(%rip), %xmm0
+; X64-NEXT: retq
+ %tmp = tail call float @llvm.copysign.f32(float %a, float 1.0)
+ ret float %tmp
+}
+
+define float @f32_neg(float %a, float %b) nounwind {
+; X86-LABEL: f32_neg:
+; X86: # %bb.0:
+; X86-NEXT: pushl %eax
+; X86-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-NEXT: orps {{\.LCPI.*}}, %xmm0
+; X86-NEXT: movss %xmm0, (%esp)
+; X86-NEXT: flds (%esp)
+; X86-NEXT: popl %eax
+; X86-NEXT: retl
+;
+; X64-LABEL: f32_neg:
+; X64: # %bb.0:
+; X64-NEXT: orps {{.*}}(%rip), %xmm0
+; X64-NEXT: retq
+ %tmp = tail call float @llvm.copysign.f32(float %a, float -1.0)
+ ret float %tmp
+}
+
+define <4 x float> @v4f32_pos(<4 x float> %a, <4 x float> %b) nounwind {
+; X86-LABEL: v4f32_pos:
+; X86: # %bb.0:
+; X86-NEXT: movaps {{.*#+}} xmm1 = [1,1,1,1]
+; X86-NEXT: andps {{\.LCPI.*}}, %xmm1
+; X86-NEXT: andps {{\.LCPI.*}}, %xmm0
+; X86-NEXT: orps %xmm1, %xmm0
+; X86-NEXT: retl
+;
+; X64-LABEL: v4f32_pos:
+; X64: # %bb.0:
+; X64-NEXT: movaps {{.*#+}} xmm1 = [1,1,1,1]
+; X64-NEXT: andps {{.*}}(%rip), %xmm1
+; X64-NEXT: andps {{.*}}(%rip), %xmm0
+; X64-NEXT: orps %xmm1, %xmm0
+; X64-NEXT: retq
+ %tmp = tail call <4 x float> @llvm.copysign.v4f32(<4 x float> %a, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>)
+ ret <4 x float> %tmp
+}
+
+define <4 x float> @v4f32_neg(<4 x float> %a, <4 x float> %b) nounwind {
+; X86-LABEL: v4f32_neg:
+; X86: # %bb.0:
+; X86-NEXT: movaps {{.*#+}} xmm1 = [-1,-1,-1,-1]
+; X86-NEXT: andps {{\.LCPI.*}}, %xmm1
+; X86-NEXT: andps {{\.LCPI.*}}, %xmm0
+; X86-NEXT: orps %xmm1, %xmm0
+; X86-NEXT: retl
+;
+; X64-LABEL: v4f32_neg:
+; X64: # %bb.0:
+; X64-NEXT: movaps {{.*#+}} xmm1 = [-1,-1,-1,-1]
+; X64-NEXT: andps {{.*}}(%rip), %xmm1
+; X64-NEXT: andps {{.*}}(%rip), %xmm0
+; X64-NEXT: orps %xmm1, %xmm0
+; X64-NEXT: retq
+ %tmp = tail call <4 x float> @llvm.copysign.v4f32(<4 x float> %a, <4 x float> <float -1.0, float -1.0, float -1.0, float -1.0>)
+ ret <4 x float> %tmp
+}
+
+declare float @llvm.copysign.f32(float, float)
+declare <4 x float> @llvm.copysign.v4f32(<4 x float>, <4 x float>)
More information about the llvm-commits
mailing list