Turn on vzeroupper insertion on call boundaries for AVX; it works as far as I know...

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@144867 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86TargetMachine.cpp b/lib/Target/X86/X86TargetMachine.cpp
index 37aa68106c7e8f991e83d4b2a158b48c02cddd2b..1c9f3bd5498ce47714ad668460ec7553ce926246 100644 (file)
--- a/lib/Target/X86/X86TargetMachine.cpp
+++ b/lib/Target/X86/X86TargetMachine.cpp
@@ -105,7 +105,7 @@ X86TargetMachine::X86TargetMachine(const Target &T, StringRef TT,
 static cl::opt<bool>
 UseVZeroUpper("x86-use-vzeroupper",
   cl::desc("Minimize AVX to SSE transition penalty"),
-  cl::init(false));
+  cl::init(true));
 
 //===----------------------------------------------------------------------===//
 // Pass Pipeline Configuration