[llvm-commits] [llvm] r147945 - in /llvm/trunk: lib/Target/X86/X86ISelDAGToDAG.cpp test/CodeGen/X86/fold-and-shift.ll

[Chandler Carruth]

Author: chandlerc
Date: Wed Jan 11 06:17:47 2012
New Revision: 147945

URL: http://llvm.org/viewvc/llvm-project?rev=147945&view=rev
Log:
Disable the transformation I added in r147936 to see if it fixes some
strange build bot failures that look like a miscompile into an infloop.
I'll investigate this tomorrow, but I'd both like to know whether my
patch is the culprit, and get the bots back to green.

Modified:
    llvm/trunk/lib/Target/X86/X86ISelDAGToDAG.cpp
    llvm/trunk/test/CodeGen/X86/fold-and-shift.ll

Modified: llvm/trunk/lib/Target/X86/X86ISelDAGToDAG.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelDAGToDAG.cpp?rev=147945&r1=147944&r2=147945&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelDAGToDAG.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelDAGToDAG.cpp Wed Jan 11 06:17:47 2012
@@ -847,6 +847,10 @@
                                     uint64_t Mask,
                                     SDValue Shift, SDValue X,
                                     X86ISelAddressMode &AM) {
+  // FIXME!! Hack to disable this and see if it is responsible for a miscompile
+  // on llvm-gcc's selfhost.
+  return true;
+
   if (Shift.getOpcode() != ISD::SRL || !Shift.hasOneUse() ||
       !isa<ConstantSDNode>(Shift.getOperand(1)))
     return true;

Modified: llvm/trunk/test/CodeGen/X86/fold-and-shift.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/fold-and-shift.ll?rev=147945&r1=147944&r2=147945&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/fold-and-shift.ll (original)
+++ llvm/trunk/test/CodeGen/X86/fold-and-shift.ll Wed Jan 11 06:17:47 2012
@@ -39,11 +39,11 @@
 ; To make matters worse, because of the two-phase zext of %i and their reuse in
 ; the function, the DAG can get confusing trying to re-use both of them and
 ; prevent easy analysis of the mask in order to match this.
-; CHECK: t3:
-; CHECK-NOT: and
-; CHECK: shrl
-; CHECK: addl (%{{...}},%{{...}},4),
-; CHECK: ret
+; BROKEN: t3:
+; BROKEN-NOT: and
+; BROKEN: shrl
+; BROKEN: addl (%{{...}},%{{...}},4),
+; BROKEN: ret
 
 entry:
   %i = load i16* %i.ptr
@@ -58,11 +58,11 @@
 define i32 @t4(i16* %i.ptr, i32* %arr) {
 ; A version of @t3 that has more zero extends and more re-use of intermediate
 ; values. This exercise slightly different bits of canonicalization.
-; CHECK: t4:
-; CHECK-NOT: and
-; CHECK: shrl
-; CHECK: addl (%{{...}},%{{...}},4),
-; CHECK: ret
+; BROKEN: t4:
+; BROKEN-NOT: and
+; BROKEN: shrl
+; BROKEN: addl (%{{...}},%{{...}},4),
+; BROKEN: ret
 
 entry:
   %i = load i16* %i.ptr