summaryrefslogtreecommitdiffstats
path: root/test/CodeGen/X86/vshift-1.ll
diff options
context:
space:
mode:
Diffstat (limited to 'test/CodeGen/X86/vshift-1.ll')
-rw-r--r--test/CodeGen/X86/vshift-1.ll22
1 files changed, 18 insertions, 4 deletions
diff --git a/test/CodeGen/X86/vshift-1.ll b/test/CodeGen/X86/vshift-1.ll
index d7a20e4..ae845e0 100644
--- a/test/CodeGen/X86/vshift-1.ll
+++ b/test/CodeGen/X86/vshift-1.ll
@@ -1,13 +1,12 @@
-; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse2 -disable-mmx -o %t -f
-; RUN: grep psllq %t | count 2
-; RUN: grep pslld %t | count 2
-; RUN: grep psllw %t | count 2
+; RUN: llc < %s -march=x86 -mattr=+sse2 -disable-mmx | FileCheck %s
; test vector shifts converted to proper SSE2 vector shifts when the shift
; amounts are the same.
define void @shift1a(<2 x i64> %val, <2 x i64>* %dst) nounwind {
entry:
+; CHECK: shift1a:
+; CHECK: psllq
%shl = shl <2 x i64> %val, < i64 32, i64 32 >
store <2 x i64> %shl, <2 x i64>* %dst
ret void
@@ -15,6 +14,9 @@ entry:
define void @shift1b(<2 x i64> %val, <2 x i64>* %dst, i64 %amt) nounwind {
entry:
+; CHECK: shift1b:
+; CHECK: movd
+; CHECK-NEXT: psllq
%0 = insertelement <2 x i64> undef, i64 %amt, i32 0
%1 = insertelement <2 x i64> %0, i64 %amt, i32 1
%shl = shl <2 x i64> %val, %1
@@ -25,6 +27,8 @@ entry:
define void @shift2a(<4 x i32> %val, <4 x i32>* %dst) nounwind {
entry:
+; CHECK: shift2a:
+; CHECK: pslld
%shl = shl <4 x i32> %val, < i32 5, i32 5, i32 5, i32 5 >
store <4 x i32> %shl, <4 x i32>* %dst
ret void
@@ -32,6 +36,9 @@ entry:
define void @shift2b(<4 x i32> %val, <4 x i32>* %dst, i32 %amt) nounwind {
entry:
+; CHECK: shift2b:
+; CHECK: movd
+; CHECK-NEXT: pslld
%0 = insertelement <4 x i32> undef, i32 %amt, i32 0
%1 = insertelement <4 x i32> %0, i32 %amt, i32 1
%2 = insertelement <4 x i32> %1, i32 %amt, i32 2
@@ -43,13 +50,20 @@ entry:
define void @shift3a(<8 x i16> %val, <8 x i16>* %dst) nounwind {
entry:
+; CHECK: shift3a:
+; CHECK: psllw
%shl = shl <8 x i16> %val, < i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5 >
store <8 x i16> %shl, <8 x i16>* %dst
ret void
}
+; Make sure the shift amount is properly zero extended.
define void @shift3b(<8 x i16> %val, <8 x i16>* %dst, i16 %amt) nounwind {
entry:
+; CHECK: shift3b:
+; CHECK: movzwl
+; CHECK: movd
+; CHECK-NEXT: psllw
%0 = insertelement <8 x i16> undef, i16 %amt, i32 0
%1 = insertelement <8 x i16> %0, i16 %amt, i32 1
%2 = insertelement <8 x i16> %0, i16 %amt, i32 2
OpenPOWER on IntegriCloud