[llvm] r318266 - [PowerPC] fix up in redundant compare elimination
Hiroshi Inoue via llvm-commits
llvm-commits at lists.llvm.org
Tue Nov 14 20:23:26 PST 2017
Author: inouehrs
Date: Tue Nov 14 20:23:26 2017
New Revision: 318266
URL: http://llvm.org/viewvc/llvm-project?rev=318266&view=rev
Log:
[PowerPC] fix up in redundant compare elimination
This patch fixes a potential problem in my previous commit (https://reviews.llvm.org/rL312514) by introducing an additional check.
Modified:
llvm/trunk/lib/Target/PowerPC/PPCMIPeephole.cpp
Modified: llvm/trunk/lib/Target/PowerPC/PPCMIPeephole.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/PPCMIPeephole.cpp?rev=318266&r1=318265&r2=318266&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/PPCMIPeephole.cpp (original)
+++ llvm/trunk/lib/Target/PowerPC/PPCMIPeephole.cpp Tue Nov 14 20:23:26 2017
@@ -814,8 +814,12 @@ static bool eligibleForCompareEliminatio
!MRI->hasOneNonDBGUse(CndReg))
return false;
- // We skip this BB if a physical register is used in comparison.
MachineInstr *CMPI = MRI->getVRegDef(CndReg);
+ // We assume compare and branch are in the same BB for ease of analysis.
+ if (CMPI->getParent() != &BB)
+ return false;
+
+ // We skip this BB if a physical register is used in comparison.
for (MachineOperand &MO : CMPI->operands())
if (MO.isReg() && !TargetRegisterInfo::isVirtualRegister(MO.getReg()))
return false;
@@ -921,7 +925,7 @@ bool PPCMIPeephole::eliminateRedundantCo
//
// As partially redundant case, we additionally handle if MBB2 has one
// additional predecessor, which has only one successor (MBB2).
- // In this case, we move the compre instruction originally in MBB2 into
+ // In this case, we move the compare instruction originally in MBB2 into
// MBBtoMoveCmp. This partially redundant case is typically appear by
// compiling a while loop; here, MBBtoMoveCmp is the loop preheader.
//
More information about the llvm-commits
mailing list