It's not necessary to do rounding for alloca operations when the requested
alignment is equal to the stack alignment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@40004 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/test/CodeGen/X86/vec_splat.ll b/test/CodeGen/X86/vec_splat.ll
new file mode 100644
index 0000000..3d73a87
--- /dev/null
+++ b/test/CodeGen/X86/vec_splat.ll
@@ -0,0 +1,22 @@
+; RUN: llvm-upgrade < %s | llvm-as | llc -march=x86 -mattr=+sse2 | grep shufps
+; RUN: llvm-upgrade < %s | llvm-as | 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
+ 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
+ store <2 x double> %tmp6, <2 x double>* %P
+ ret void
+}