summaryrefslogtreecommitdiffstats
path: root/llvm/test/CodeGen/X86/vec_splat.ll
diff options
context:
space:
mode:
authorEvan Cheng <evan.cheng@apple.com>2008-04-05 00:30:36 +0000
committerEvan Cheng <evan.cheng@apple.com>2008-04-05 00:30:36 +0000
commitf77b5ef3d054a78bc8ab3a61742112eb7faea1f4 (patch)
treed282cabee197e54a06877477d9bbda5fcdd65b5e /llvm/test/CodeGen/X86/vec_splat.ll
parent0ce4a7cc4440f1f66d91ad61057790eb196f2e4a (diff)
downloadbcm5719-llvm-f77b5ef3d054a78bc8ab3a61742112eb7faea1f4.tar.gz
bcm5719-llvm-f77b5ef3d054a78bc8ab3a61742112eb7faea1f4.zip
Favors pshufd over shufps when shuffling elements from one vector. pshufd is faster than shufps.
llvm-svn: 49244
Diffstat (limited to 'llvm/test/CodeGen/X86/vec_splat.ll')
-rw-r--r--llvm/test/CodeGen/X86/vec_splat.ll6
1 files changed, 3 insertions, 3 deletions
diff --git a/llvm/test/CodeGen/X86/vec_splat.ll b/llvm/test/CodeGen/X86/vec_splat.ll
index c6100ecd094..64222e40ff2 100644
--- a/llvm/test/CodeGen/X86/vec_splat.ll
+++ b/llvm/test/CodeGen/X86/vec_splat.ll
@@ -1,7 +1,7 @@
-; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse2 | grep shufps
+; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse2 | grep pshufd
; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse3 | grep movddup
-define void @test_v4sf(<4 x float>* %P, <4 x float>* %Q, float %X) {
+define void @test_v4sf(<4 x float>* %P, <4 x float>* %Q, float %X) nounwind {
%tmp = insertelement <4 x float> zeroinitializer, float %X, i32 0 ; <<4 x float>> [#uses=1]
%tmp2 = insertelement <4 x float> %tmp, float %X, i32 1 ; <<4 x float>> [#uses=1]
%tmp4 = insertelement <4 x float> %tmp2, float %X, i32 2 ; <<4 x float>> [#uses=1]
@@ -12,7 +12,7 @@ define void @test_v4sf(<4 x float>* %P, <4 x float>* %Q, float %X) {
ret void
}
-define void @test_v2sd(<2 x double>* %P, <2 x double>* %Q, double %X) {
+define void @test_v2sd(<2 x double>* %P, <2 x double>* %Q, double %X) nounwind {
%tmp = insertelement <2 x double> zeroinitializer, double %X, i32 0 ; <<2 x double>> [#uses=1]
%tmp2 = insertelement <2 x double> %tmp, double %X, i32 1 ; <<2 x double>> [#uses=1]
%tmp4 = load <2 x double>* %Q ; <<2 x double>> [#uses=1]
OpenPOWER on IntegriCloud