Use movups to lower memcpy and memset even if it's not fast (like corei7).
The theory is it's still faster than a pair of movq / a quad of movl. This
will probably hurt older chips like P4 but should run faster on current
and future Intel processors. rdar://8817010
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@122955 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/test/CodeGen/X86/small-byval-memcpy.ll b/test/CodeGen/X86/small-byval-memcpy.ll
index 1b596b5..8f69b11 100644
--- a/test/CodeGen/X86/small-byval-memcpy.ll
+++ b/test/CodeGen/X86/small-byval-memcpy.ll
@@ -1,8 +1,12 @@
-; RUN: llc < %s -mtriple=i386-apple-darwin -mcpu=core2 | grep movsd | count 8
-; RUN: llc < %s -mtriple=i386-apple-darwin -mcpu=nehalem | grep movups | count 2
+; RUN: llc < %s -mtriple=i386-apple-darwin -mcpu=nehalem | FileCheck %s
define void @ccosl({ x86_fp80, x86_fp80 }* noalias sret %agg.result, { x86_fp80, x86_fp80 }* byval align 4 %z) nounwind {
entry:
+; CHECK: ccosl:
+; CHECK: movaps
+; CHECK: movaps
+; CHECK: movups
+; CHECK: movups
%iz = alloca { x86_fp80, x86_fp80 } ; <{ x86_fp80, x86_fp80 }*> [#uses=3]
%tmp1 = getelementptr { x86_fp80, x86_fp80 }* %z, i32 0, i32 1 ; <x86_fp80*> [#uses=1]
%tmp2 = load x86_fp80* %tmp1, align 16 ; <x86_fp80> [#uses=1]