Implement PR3495: local spiller optimization. The local spiller can now keep availabi...

[oota-llvm.git] / test / CodeGen / X86 / vec_splat.ll

diff --git a/test/CodeGen/X86/vec_splat.ll b/test/CodeGen/X86/vec_splat.ll
index b50038c5861a64932d4159f01cfb5df076ba2929..64222e40ff23bf0efa25263b1ca43a9cdda2296d 100644 (file)
--- a/test/CodeGen/X86/vec_splat.ll
+++ b/test/CodeGen/X86/vec_splat.ll
@@ -1,22 +1,22 @@
-; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse2 | grep shufps &&
-; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse2 | grep movddup
+; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse2 | grep pshufd
+; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse3 | grep movddup
 
-void %test_v4sf(<4 x float>* %P, <4 x float>* %Q, float %X) {
-       %tmp = insertelement <4 x float> zeroinitializer, float %X, uint 0
-       %tmp2 = insertelement <4 x float> %tmp, float %X, uint 1
-       %tmp4 = insertelement <4 x float> %tmp2, float %X, uint 2
-       %tmp6 = insertelement <4 x float> %tmp4, float %X, uint 3
-       %tmp8 = load <4 x float>* %Q
-       %tmp10 = mul <4 x float> %tmp8, %tmp6
+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]
+       %tmp6 = insertelement <4 x float> %tmp4, float %X, i32 3                ; <<4 x float>> [#uses=1]
+       %tmp8 = load <4 x float>* %Q            ; <<4 x float>> [#uses=1]
+       %tmp10 = mul <4 x float> %tmp8, %tmp6           ; <<4 x float>> [#uses=1]
        store <4 x float> %tmp10, <4 x float>* %P
        ret void
 }
 
-void %test_v2sd(<2 x double>* %P, <2 x double>* %Q, double %X) {
-       %tmp = insertelement <2 x double> zeroinitializer, double %X, uint 0
-       %tmp2 = insertelement <2 x double> %tmp, double %X, uint 1
-       %tmp4 = load <2 x double>* %Q
-       %tmp6 = mul <2 x double> %tmp4, %tmp2
+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]
+       %tmp6 = mul <2 x double> %tmp4, %tmp2           ; <<2 x double>> [#uses=1]
        store <2 x double> %tmp6, <2 x double>* %P
        ret void
 }