mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2026-04-25 05:22:04 +00:00
Allow MachineCSE to coalesce trivial subregister copies the same way that it coalesces normal copies.
Without this, MachineCSE is powerless to handle redundant operations with truncated source operands.
This required fixing the 2-addr pass to handle tied subregisters. It isn't clear what combinations of subregisters can legally be tied, but the simple case of truncated source operands is now safely handled:
%vreg11<def> = COPY %vreg1:sub_32bit; GR32:%vreg11 GR64:%vreg1
%vreg12<def> = COPY %vreg2:sub_32bit; GR32:%vreg12 GR64:%vreg2
%vreg13<def,tied1> = ADD32rr %vreg11<tied0>, %vreg12<kill>, %EFLAGS<imp-def>
Test case: cse-add-with-overflow.ll.
This exposed an existing bug in
PPCInstrInfo::commuteInstruction. Thanks to Rafael for the test case:
PowerPC/crash.ll.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@197465 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
@@ -131,13 +131,18 @@ bool MachineCSE::PerformTrivialCoalescing(MachineInstr *MI,
|
||||
unsigned SrcReg = DefMI->getOperand(1).getReg();
|
||||
if (!TargetRegisterInfo::isVirtualRegister(SrcReg))
|
||||
continue;
|
||||
if (DefMI->getOperand(0).getSubReg() || DefMI->getOperand(1).getSubReg())
|
||||
if (DefMI->getOperand(0).getSubReg())
|
||||
continue;
|
||||
if (!MRI->constrainRegClass(SrcReg, MRI->getRegClass(Reg)))
|
||||
unsigned SrcSubReg = DefMI->getOperand(1).getSubReg();
|
||||
const TargetRegisterClass *RC = MRI->getRegClass(Reg);
|
||||
if (SrcSubReg)
|
||||
RC = TRI->getMatchingSuperRegClass(MRI->getRegClass(SrcReg), RC,
|
||||
SrcSubReg);
|
||||
if (!MRI->constrainRegClass(SrcReg, RC))
|
||||
continue;
|
||||
DEBUG(dbgs() << "Coalescing: " << *DefMI);
|
||||
DEBUG(dbgs() << "*** to: " << *MI);
|
||||
MO.setReg(SrcReg);
|
||||
MO.substVirtReg(SrcReg, SrcSubReg, *TRI);
|
||||
MRI->clearKillFlags(SrcReg);
|
||||
DefMI->eraseFromParent();
|
||||
++NumCoalesces;
|
||||
|
||||
Reference in New Issue
Block a user