4 files changed, 28 insertions, 11 deletions
diff --git a/test/CodeGen/ARM/vstlane.ll b/test/CodeGen/ARM/vstlane.ll
index d1bc15a..08b7232 100644
--- a/test/CodeGen/ARM/vstlane.ll
+++ b/test/CodeGen/ARM/vstlane.ll
@@ -54,7 +54,8 @@ define void @vst1lanef(float* %A, <2 x float>* %B) nounwind {
 
 define void @vst1laneQi8(i8* %A, <16 x i8>* %B) nounwind {
 ;CHECK: vst1laneQi8:
-;CHECK: vst1.8 {d17[1]}, [r0]
+; // Can use scalar load. No need to use vectors.
+; // CHE-CK: vst1.8 {d17[1]}, [r0]
 	%tmp1 = load <16 x i8>* %B
         %tmp2 = extractelement <16 x i8> %tmp1, i32 9
         store i8 %tmp2, i8* %A, align 8
@@ -72,7 +73,8 @@ define void @vst1laneQi16(i16* %A, <8 x i16>* %B) nounwind {
 
 define void @vst1laneQi32(i32* %A, <4 x i32>* %B) nounwind {
 ;CHECK: vst1laneQi32:
-;CHECK: vst1.32 {d17[1]}, [r0, :32]
+; // Can use scalar load. No need to use vectors.
+; // CHE-CK: vst1.32 {d17[1]}, [r0, :32]
 	%tmp1 = load <4 x i32>* %B
         %tmp2 = extractelement <4 x i32> %tmp1, i32 3
         store i32 %tmp2, i32* %A, align 8
@@ -82,7 +84,8 @@ define void @vst1laneQi32(i32* %A, <4 x i32>* %B) nounwind {
 ;Check for a post-increment updating store.
 define void @vst1laneQi32_update(i32** %ptr, <4 x i32>* %B) nounwind {
 ;CHECK: vst1laneQi32_update:
-;CHECK: vst1.32 {d17[1]}, [r1, :32]!
+; // Can use scalar load. No need to use vectors.
+; // CHE-CK: vst1.32 {d17[1]}, [r1, :32]!
 	%A = load i32** %ptr
 	%tmp1 = load <4 x i32>* %B
 	%tmp2 = extractelement <4 x i32> %tmp1, i32 3
@@ -94,7 +97,8 @@ define void @vst1laneQi32_update(i32** %ptr, <4 x i32>* %B) nounwind {
 
 define void @vst1laneQf(float* %A, <4 x float>* %B) nounwind {
 ;CHECK: vst1laneQf:
-;CHECK: vst1.32 {d17[1]}, [r0]
+; // Can use scalar load. No need to use vectors.
+; // CHE-CK: vst1.32 {d17[1]}, [r0]
 	%tmp1 = load <4 x float>* %B
         %tmp2 = extractelement <4 x float> %tmp1, i32 3
         store float %tmp2, float* %A
diff --git a/test/CodeGen/X86/2011-05-09-loaduse.ll b/test/CodeGen/X86/2011-05-09-loaduse.ll
new file mode 100644
index 0000000..8673d74
--- /dev/null
+++ b/test/CodeGen/X86/2011-05-09-loaduse.ll
@@ -0,0 +1,13 @@
+; RUN: llc < %s -march=x86 -mcpu=corei7 | FileCheck %s
+
+;CHECK: test
+;CHECK-not: pshufd
+;CHECK: ret
+define float @test(<4 x float>* %A) nounwind {
+entry:
+  %T = load <4 x float>* %A
+  %R = extractelement <4 x float> %T, i32 3
+  store <4 x float><float 0.0, float 0.0, float 0.0, float 0.0>, <4 x float>* %A
+  ret float %R
+}
+
diff --git a/test/CodeGen/X86/vec_extract-sse4.ll b/test/CodeGen/X86/vec_extract-sse4.ll
index dab5dd1..f487654 100644
--- a/test/CodeGen/X86/vec_extract-sse4.ll
+++ b/test/CodeGen/X86/vec_extract-sse4.ll
@@ -1,8 +1,8 @@
-; RUN: llc < %s -march=x86 -mattr=+sse41 -o %t
-; RUN: grep extractps   %t | count 1
-; RUN: grep pextrd      %t | count 1
+; RUN: llc < %s -mcpu=corei7 -march=x86 -mattr=+sse41 -o %t
+; RUN: not grep extractps   %t
+; RUN: not grep pextrd      %t
 ; RUN: not grep pshufd  %t
-; RUN: not grep movss   %t
+; RUN: grep movss   %t | count 2
 
 define void @t1(float* %R, <4 x float>* %P1) nounwind {
 	%X = load <4 x float>* %P1
diff --git a/test/CodeGen/X86/vec_extract.ll b/test/CodeGen/X86/vec_extract.ll
index b013730..2c8796b 100644
--- a/test/CodeGen/X86/vec_extract.ll
+++ b/test/CodeGen/X86/vec_extract.ll
@@ -1,7 +1,7 @@
-; RUN: llc < %s -march=x86 -mattr=+sse2,-sse41 -o %t
-; RUN: grep movss    %t | count 3
+; RUN: llc < %s -mcpu=corei7 -march=x86 -mattr=+sse2,-sse41 -o %t
+; RUN: grep movss    %t | count 4
 ; RUN: grep movhlps  %t | count 1
-; RUN: grep pshufd   %t | count 1
+; RUN: not grep pshufd   %t 
 ; RUN: grep unpckhpd %t | count 1
 
 define void @test1(<4 x float>* %F, float* %f) nounwind {