[llvm-commits] [llvm] r134664 - in /llvm/trunk: lib/Target/X86/X86ISelLowering.cpp test/CodeGen/X86/membarrier.ll

[Eric Christopher]

Author: echristo
Date: Thu Jul  7 19:04:56 2011
New Revision: 134664

URL: http://llvm.org/viewvc/llvm-project?rev=134664&view=rev
Log:
Go ahead and emit the barrier on x86-64 even without sse2. The
processor supports it just fine.

Fixes PR9675 and rdar://9740801

Added:
    llvm/trunk/test/CodeGen/X86/membarrier.ll
Modified:
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=134664&r1=134663&r2=134664&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Thu Jul  7 19:04:56 2011
@@ -9067,10 +9067,11 @@
 SDValue X86TargetLowering::LowerMEMBARRIER(SDValue Op, SelectionDAG &DAG) const{
   DebugLoc dl = Op.getDebugLoc();
 
-  if (!Subtarget->hasSSE2()) {
+  // Go ahead and emit the fence on x86-64 even if we asked for no-sse2.
+  // There isn't any reason to disable it if the target processor supports it.
+  if (!Subtarget->hasSSE2() && !Subtarget->is64Bit()) {
     SDValue Chain = Op.getOperand(0);
-    SDValue Zero = DAG.getConstant(0,
-                                   Subtarget->is64Bit() ? MVT::i64 : MVT::i32);
+    SDValue Zero = DAG.getConstant(0, MVT::i32);
     SDValue Ops[] = {
       DAG.getRegister(X86::ESP, MVT::i32), // Base
       DAG.getTargetConstant(1, MVT::i8),   // Scale

Added: llvm/trunk/test/CodeGen/X86/membarrier.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/membarrier.ll?rev=134664&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/membarrier.ll (added)
+++ llvm/trunk/test/CodeGen/X86/membarrier.ll Thu Jul  7 19:04:56 2011
@@ -0,0 +1,15 @@
+; RUN: llc < %s -march=x86-64 -mattr=-sse -O0
+; PR9675
+
+define i32 @t() {
+entry:
+  %i = alloca i32, align 4
+  store i32 1, i32* %i, align 4
+  call void @llvm.memory.barrier(i1 true, i1 true, i1 true, i1 true, i1 true)
+  %0 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %i, i32 1)
+  call void @llvm.memory.barrier(i1 true, i1 true, i1 true, i1 true, i1 true)
+  ret i32 0
+}
+
+declare i32 @llvm.atomic.load.sub.i32.p0i32(i32* nocapture, i32) nounwind
+declare void @llvm.memory.barrier(i1, i1, i1, i1, i1) nounwind