CodeGen peephole: fold redundant phys reg copies

[oota-llvm.git] / test / CodeGen / X86 / 2011-12-15-vec_shift.ll

diff --git a/test/CodeGen/X86/2011-12-15-vec_shift.ll b/test/CodeGen/X86/2011-12-15-vec_shift.ll
index dc3a08bb4daf5d3f92f0e6e3cd4e9fd0db7451ae..4d49b3af88ee1cccc9020ae6e4686a0700068b39 100644 (file)
--- a/test/CodeGen/X86/2011-12-15-vec_shift.ll
+++ b/test/CodeGen/X86/2011-12-15-vec_shift.ll
@@ -1,5 +1,5 @@
-; RUN: llc -march=x86-64 -mattr=+sse41 -mcpu=penryn < %s | FileCheck %s -check-prefix=CHECK-W-SSE4
-; RUN: llc -march=x86-64 -mattr=-sse41 -mcpu=penryn < %s | FileCheck %s -check-prefix=CHECK-WO-SSE4
+; RUN: llc -march=x86-64 -mattr=+sse4.1 -mcpu=penryn < %s | FileCheck %s -check-prefix=CHECK-W-SSE4
+; RUN: llc -march=x86-64 -mattr=-sse4.1 -mcpu=penryn < %s | FileCheck %s -check-prefix=CHECK-WO-SSE4
 ; Test case for r146671
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.7"
@@ -12,8 +12,8 @@ define <16 x i8> @shift(<16 x i8> %a, <16 x i8> %b) nounwind {
 
   ; Make sure we're masking and pcmp'ing the VSELECT conditon vector.
   ; CHECK-WO-SSE4: psllw $5, [[REG1:%xmm.]]
-  ; CHECK-WO-SSE4: pand [[REG1]], [[REG2:%xmm.]]
-  ; CHECK-WO-SSE4: pcmpeqb {{%xmm., }}[[REG2]]
+  ; CHECK-WO-SSE4: pxor [[REG2:%xmm.]], [[REG2:%xmm.]]
+  ; CHECK-WO-SSE4: pcmpgtb {{%xmm., }}[[REG2]]
   %1 = shl <16 x i8> %a, %b
   ret <16 x i8> %1
 }