[x86] Fix an embarressing bug in the INSERTPS formation code. The mask

[oota-llvm.git] / lib / Target / X86 / X86VZeroUpper.cpp

diff --git a/lib/Target/X86/X86VZeroUpper.cpp b/lib/Target/X86/X86VZeroUpper.cpp
index d4341b9fc35aa2870fc033e91c45ec31b201ce55..d93baeb7002e3da51954c29db41991143968fac5 100644 (file)
--- a/lib/Target/X86/X86VZeroUpper.cpp
+++ b/lib/Target/X86/X86VZeroUpper.cpp
@@ -14,7 +14,6 @@
 //
 //===----------------------------------------------------------------------===//
 
-#define DEBUG_TYPE "x86-vzeroupper"
 #include "X86.h"
 #include "X86InstrInfo.h"
 #include "X86Subtarget.h"
@@ -28,6 +27,8 @@
 #include "llvm/Target/TargetInstrInfo.h"
 using namespace llvm;
 
+#define DEBUG_TYPE "x86-vzeroupper"
+
 STATISTIC(NumVZU, "Number of vzeroupper instructions inserted");
 
 namespace {
@@ -246,9 +247,10 @@ void VZeroUpperInserter::processBasicBlock(MachineBasicBlock &MBB) {
 /// runOnMachineFunction - Loop over all of the basic blocks, inserting
 /// vzero upper instructions before function calls.
 bool VZeroUpperInserter::runOnMachineFunction(MachineFunction &MF) {
-  if (MF.getTarget().getSubtarget<X86Subtarget>().hasAVX512())
+  const X86Subtarget &ST = MF.getTarget().getSubtarget<X86Subtarget>();
+  if (!ST.hasAVX() || ST.hasAVX512())
     return false;
-  TII = MF.getTarget().getInstrInfo();
+  TII = MF.getSubtarget().getInstrInfo();
   MachineRegisterInfo &MRI = MF.getRegInfo();
   EverMadeChange = false;