summaryrefslogtreecommitdiff
path: root/lib/CodeGen/MachineLICM.cpp
AgeCommit message (Expand)AuthorFilesLines
2009-07-25More migration to raw_ostream, the water has dried up around the iostream hole.Daniel Dunbar1-6/+7
2009-02-27MachineLICM CSE should match destination register classes; avoid hoisting imp...Evan Cheng1-3/+13
2009-02-05Machine LICM increases register pressure and it almost always increase code s...Evan Cheng1-0/+4
2009-02-05Teach machine licm to CSE hoisted instructions.Evan Cheng1-23/+88
2009-02-04Don't call isInvariantLoad twice.Evan Cheng1-9/+2
2009-02-04Don't bother hoisting out a "cheap" instruction if all of its uses are PHIs. ...Evan Cheng1-10/+55
2009-02-04For now, only hoist re-materilizable instructions. LICM will increase registe...Evan Cheng1-2/+11
2009-01-15Simplify the MachineLICM pass by having it only traverse outerDan Gohman1-122/+60
2009-01-05Tidy up #includes, deleting a bunch of unnecessary #includes.Dan Gohman1-1/+0
2008-12-23Use isTerminator() instead of isBranch()||isReturn() inDan Gohman1-1/+1
2008-12-18Give MachineLICM a name, for -time-passes etc.Dan Gohman1-0/+2
2008-10-03Switch the MachineOperand accessors back to the short names likeDan Gohman1-1/+1
2008-09-04Tidy up several unbeseeming casts from pointer to intptr_t.Dan Gohman1-1/+1
2008-08-31Cosmetic changes to Machine LICM. No functionality change.Bill Wendling1-31/+36
2008-08-20Don't hoist instructions that define a physical register.Bill Wendling1-1/+8
2008-08-18Don't hoist instructions that have implicit uses or defines. Before, it wasBill Wendling1-0/+5
2008-07-07Prevent option name conflict.Bill Wendling1-1/+1
2008-05-13Clean up the use of static and anonymous namespaces. This turned upDan Gohman1-4/+4
2008-05-12Revert the previous commit. Go ahead and hoist rematerializable instructions.Bill Wendling1-4/+0
2008-05-12One real change - don't hoist something that's trivially rematerializable. It'sBill Wendling1-18/+19
2008-03-10Minor cleanup. No functionality change.Bill Wendling1-6/+8
2008-02-26Rename PrintableName to Name.Bill Wendling1-2/+2
2008-02-26Change "Name" to "AsmName" in the target register info. Gee, a refactoring toolBill Wendling1-2/+2
2008-02-10Rename MRegisterInfo to TargetRegisterInfo.Dan Gohman1-6/+6
2008-01-10Simplify the side effect stuff a bit more and make licm/sinkingChris Lattner1-5/+22
2008-01-07rename TargetInstrDescriptor -> TargetInstrDesc.Chris Lattner1-4/+4
2008-01-07Rename MachineInstr::getInstrDescriptor -> getDesc(), which reflectsChris Lattner1-8/+6
2008-01-04Add that this preserves some analyses.Bill Wendling1-1/+3
2008-01-04Move option to enable machine LICM into LLVMTargetMachine.cpp.Bill Wendling1-10/+0
2008-01-04Call the parent's getAnalysisUsage.Bill Wendling1-0/+1
2008-01-04remove dead #includes and reorder the rest.Chris Lattner1-10/+6
2008-01-02Use the correct MachineRegisterInfo object.Bill Wendling1-2/+1
2008-01-02Remove dead code.Bill Wendling1-35/+0
2008-01-02Use the new architecture to get the containing machine basic block for a machineBill Wendling1-28/+28
2007-12-29Remove attribution from file headers, per discussion on llvmdev.Chris Lattner1-2/+2
2007-12-20Updated comments to reflect what "side effects" means in this situation.Bill Wendling1-10/+3
2007-12-18Add debugging info. Use the newly created "hasUnmodelledSideEffects" method.Bill Wendling1-21/+38
2007-12-11Need to grow the indexed map. Added debug statements.Bill Wendling1-21/+27
2007-12-11Simplify slightly.Bill Wendling1-3/+4
2007-12-11Blark! How in the world did this work without this?!Bill Wendling1-4/+9
2007-12-11- Update the virtual reg to machine instruction map when hoisting.Bill Wendling1-2/+12
2007-12-11Checking for "zero operands" during the "CanHoistInst()" method isn't necessaryBill Wendling1-11/+4
2007-12-08Reverting 44702. It wasn't correct to rename them.Bill Wendling1-13/+13
2007-12-08Renaming:Bill Wendling1-1/+1
2007-12-08Incorporated comments from Evan and Chris:Bill Wendling1-58/+54
2007-12-07Initial commit of the machine code LICM pass. It successfully hoists this:Bill Wendling1-0/+336