summaryrefslogtreecommitdiffstats
path: root/llvm/lib/CodeGen/TailDuplication.cpp
Commit message (Collapse)AuthorAgeFilesLines
* Move most of the pre BB code to TailDuplicateAndUpdate. Change theRafael Espindola2011-07-041-112/+125
| | | | | | HasIndirectbr variable to be just that. No functionality change. llvm-svn: 134371
* Reduce indentation and fix the count of how many PHIs we have inserted.Rafael Espindola2011-07-041-75/+80
| | | | llvm-svn: 134370
* - Rename TargetInstrDesc, TargetOperandInfo to MCInstrDesc and MCOperandInfo andEvan Cheng2011-06-281-2/+2
| | | | | | | | sink them into MC layer. - Added MCInstrInfo, which captures the tablegen generated static data. Chang TargetInstrInfo so it's based off MCInstrInfo. llvm-svn: 134021
* SimplifyRafael Espindola2011-06-241-14/+6
| | | | llvm-svn: 133798
* Now that bb with phis are not considered simple, duplicate them even ifRafael Espindola2011-06-241-12/+18
| | | | | | we cannot duplicate to every predecessor. llvm-svn: 133797
* Simplify now that blocks with phis are not considered simple.Rafael Espindola2011-06-241-15/+0
| | | | llvm-svn: 133793
* Move more logic to shouldTailDuplicate and only duplicate regular bb beforeRafael Espindola2011-06-231-21/+45
| | | | | | | | | | register allocation if it has a indirectbr or if we can duplicate it to every predecessor. This fixes the SingleSource/Benchmarks/Shootout-C++/matrix.cpp regression but keeps the previous improvements to sunspider. llvm-svn: 133682
* Reenable tail duplication of bb with just an unconditional jump, butRafael Espindola2011-06-221-4/+4
| | | | | | don't remove blocks that have their address taken. llvm-svn: 133659
* Revert r133607. This is causing failures in the Clang gccTestSuite.Chad Rosier2011-06-221-3/+3
| | | | | | Specifically, gcc.c-torture/compile/pr21356.c. llvm-svn: 133646
* Reenable the optimization added in 133415, but change the definition of a ↵Rafael Espindola2011-06-221-3/+3
| | | | | | | | | "simple" bb to be one with only one unconditional branch and no phis. Duplicating the phis in this case is possible, but requeres liveness analysis or breaking edges. llvm-svn: 133607
* Disable again.Rafael Espindola2011-06-201-1/+1
| | | | llvm-svn: 133446
* Re enable 133415 with two fixesRafael Espindola2011-06-201-3/+8
| | | | | | | | * Don't introduce a duplicated bb in the CFG * When making a branch unconditional, clear the PredCond array so that it is really unconditional. llvm-svn: 133432
* Disable the logic added by rafael in commit 133415 to see if it brings theDuncan Sands2011-06-201-1/+1
| | | | | | | | dragonegg buildbots back to life. Original commit message: Teach early dup how to duplicate basic blocks with one successor and only phi instructions into more complex blocks. llvm-svn: 133430
* Fix MSVC build. next() function already exists in the MSVC headers. This ↵Francois Pichet2011-06-201-1/+1
| | | | | | create a overload conflict. Make sure we pick up the llvm one. llvm-svn: 133416
* Teach early dup how to duplicate basic blocks with one successor and only ↵Rafael Espindola2011-06-201-2/+142
| | | | | | | | phi instructions into more complex blocks. llvm-svn: 133415
* Two fixes relating to debug value:Rafael Espindola2011-06-171-0/+10
| | | | | | | | * We should change the generated code because of a debug use. * Avoid creating debug uses of undef, as they become a kill. Test to follow. llvm-svn: 133255
* Enable early duplication of small blocks. There are still improvements toRafael Espindola2011-06-171-29/+32
| | | | | | be made, but this is already a win. llvm-svn: 133240
* Removed tabs. Also fixed my editor...Rafael Espindola2011-06-101-3/+2
| | | | llvm-svn: 132857
* Remove duplicated test.Rafael Espindola2011-06-101-4/+3
| | | | | | Thanks Bob Wilson for noticing it! llvm-svn: 132851
* Make the optional verification step more strict.Rafael Espindola2011-06-091-1/+8
| | | | llvm-svn: 132822
* Avoid a gcc warning about multiline comments.Rafael Espindola2011-06-091-6/+6
| | | | llvm-svn: 132821
* On last fix to the early tail duplication.Rafael Espindola2011-06-091-4/+60
| | | | | | | | With this I am able to bootstrap clang with early tail duplication enabled for any small bb and setting tail-dup-size to a relatively large value(8) to stress this code. llvm-svn: 132816
* Also consider phi nodes when deciding if a register is live out.Rafael Espindola2011-06-091-10/+34
| | | | llvm-svn: 132814
* AnalyzeBranch modifies the bb, but we don't want to modify a bb withRafael Espindola2011-06-091-7/+6
| | | | | | eh edges. Swap the order of the checks to avoid it. llvm-svn: 132806
* A PHI in this basic block is a use in another basic block.Rafael Espindola2011-06-091-1/+1
| | | | llvm-svn: 132805
* Refactor some checks into shouldTailDuplicate. Update comments.Rafael Espindola2011-06-091-18/+29
| | | | | | No functionality change. llvm-svn: 132798
* Fix count.Rafael Espindola2011-06-081-1/+1
| | | | llvm-svn: 132749
* Count how many phis we are creating.Rafael Espindola2011-06-081-0/+2
| | | | llvm-svn: 132748
* Update comments.Evan Cheng2011-02-041-2/+3
| | | | llvm-svn: 124843
* Respect the -tail-dup-size command line option even when optimizing for size.Jakob Stoklund Olesen2011-01-301-1/+2
| | | | | | | This is similar to the -unroll-threshold option. There should be no change in behavior when -tail-dup-size is not explicit on the llc command line. llvm-svn: 124564
* Re-apply r124518 with fix. Watch out for invalidated iterator.Evan Cheng2011-01-291-4/+7
| | | | llvm-svn: 124526
* Revert r124518. It broke Linux self-host.Evan Cheng2011-01-291-7/+4
| | | | llvm-svn: 124522
* Re-commit r124462 with fixes. Tail recursion elim will now dup ret into ↵Evan Cheng2011-01-291-4/+7
| | | | | | unconditional predecessor to enable TCE on demand. llvm-svn: 124518
* Revert r124462. There are a few big regressions that I need to fix first.Evan Cheng2011-01-281-7/+4
| | | | llvm-svn: 124478
* - Stop simplifycfg from duplicating "ret" instructions into unconditionalEvan Cheng2011-01-281-4/+7
| | | | | | | | branches. PR8575, rdar://5134905, rdar://8911460. - Allow codegen tail duplication to dup small return blocks after register allocation is done. llvm-svn: 124462
* Simplify a bunch of isVirtualRegister() and isPhysicalRegister() logic.Jakob Stoklund Olesen2011-01-101-1/+1
| | | | | | | | These functions not longer assert when passed 0, but simply return false instead. No functional change intended. llvm-svn: 123155
* Reapply r110396, with fixes to appease the Linux buildbot gods.Owen Anderson2010-08-061-1/+1
| | | | llvm-svn: 110460
* Revert r110396 to fix buildbots.Owen Anderson2010-08-061-1/+1
| | | | llvm-svn: 110410
* Don't use PassInfo* as a type identifier for passes. Instead, use the ↵Owen Anderson2010-08-051-1/+1
| | | | | | | | address of the static ID member as the sole unique type identifier. Clean up APIs related to this change. llvm-svn: 110396
* Remove many calls to TII::isMoveInstr. Targets should be producing COPY anyway.Jakob Stoklund Olesen2010-07-161-8/+9
| | | | | | TII::isMoveInstr is going tobe completely removed. llvm-svn: 108507
* Replace copyRegToReg with COPY everywhere in lib/CodeGen except for FastISel.Jakob Stoklund Olesen2010-07-101-10/+8
| | | | llvm-svn: 108062
* Add a DebugLoc argument to TargetInstrInfo::copyRegToReg, so that itDan Gohman2010-05-061-2/+2
| | | | | | doesn't have to guess. llvm-svn: 103194
* Ignore debug value instructions while analyzing BB for tail duplication.Devang Patel2010-03-161-1/+1
| | | | llvm-svn: 98675
* eliminate InvalidateLabel and LabelIDList from MMI and replaceChris Lattner2010-03-141-11/+0
| | | | | | them with a counter. llvm-svn: 98462
* Reuse operand location when updating PHI instructions.Jakob Stoklund Olesen2010-02-111-7/+26
| | | | | | | | Calling RemoveOperand is very expensive on huge PHI instructions. This makes early tail duplication run twice as fast on the Firefox JavaScript interpreter. llvm-svn: 95832
* move target-independent opcodes out of TargetInstrInfoChris Lattner2010-02-091-5/+5
| | | | | | | | | into TargetOpcodes.h. #include the new TargetOpcodes.h into MachineInstr. Add new inline accessors (like isPHI()) to MachineInstr, and start using them throughout the codebase. llvm-svn: 95687
* Treat indirect branches specially only during pre-regalloc tail duplication,Bob Wilson2010-01-161-13/+13
| | | | | | not during the later post-alloc tail duplication. llvm-svn: 93600
* Simplify logic. Any functional change is unintended.Jakob Stoklund Olesen2010-01-151-6/+2
| | | | llvm-svn: 93540
* Change pre-regalloc tail duplication to only duplicate indirect branch blocks.Bob Wilson2010-01-151-4/+11
| | | | | | | | | The pre-regalloc pass caused some regressions in both compile time and performance of the generated code, and it did not improve performance, except for indirect branches. I also moved the check for single-block loops to speed up the common case when running the taildup pass before reg allocation. llvm-svn: 93505
* Add Target hook to duplicate machine instructions.Jakob Stoklund Olesen2010-01-061-1/+1
| | | | | | | Some instructions refer to unique labels, and so cannot be trivially cloned with CloneMachineInstr. llvm-svn: 92873
OpenPOWER on IntegriCloud