summaryrefslogtreecommitdiffstats
path: root/llvm/lib
Commit message (Collapse)AuthorAgeFilesLines
* Add a "nop filler" pass to SPU.Kalle Raiskila2011-01-116-1/+166
| | | | | | | | | | | | | | | Filling no-ops is done just before emitting of assembly, when the instruction stream is final. No-ops are inserted to align the instructions so the dual-issue of the pipeline is utilized. This speeds up generated code with a minimum of 1% on a select set of algorithms. This pass may be redundant if the instruction scheduler and all subsequent passes that modify the instruction stream (prolog+epilog inserter, register scavenger, are there others?) are made aware of the instruction alignments. llvm-svn: 123226
* Temporarily revert 123133, it's causing some regressions and I'm tryingEric Christopher2011-01-111-8/+4
| | | | | | to get a testcase. llvm-svn: 123225
* update memdep when an instruction is deleted. This code isn'tChris Lattner2011-01-111-2/+5
| | | | | | | actually reached in the testcase in PR8954, but it's safe and good practice. llvm-svn: 123224
* when MergeBlockIntoPredecessor merges two blocks, update MemDep if itChris Lattner2011-01-111-0/+4
| | | | | | is floating around in the ether. llvm-svn: 123223
* Fix FoldSingleEntryPHINodes to update memdep and AA when it deletesChris Lattner2011-01-112-5/+21
| | | | | | | | | | phi nodes. It is called from MergeBlockIntoPredecessor which is called from GVN, which claims to preserve these. I'm skeptical that this is the actual problem behind PR8954, but this is a stab in the right direction. llvm-svn: 123222
* random cleanupsChris Lattner2011-01-112-3/+2
| | | | llvm-svn: 123221
* remove a bogus assertion: the latch block of a loop is not Chris Lattner2011-01-111-6/+5
| | | | | | | neccesarily an uncond branch to the header. This fixes PR8955 (the assertion tripping). llvm-svn: 123219
* the GEP faq says that only inbounds geps are guaranteed to not overflow.Chris Lattner2011-01-111-2/+3
| | | | llvm-svn: 123218
* Revert r123207: "Turn on memdep's verifyRemoved() in an attempt to smoke out ↵Jakob Stoklund Olesen2011-01-111-3/+1
| | | | | | | | the cause of our gcc bootstrap miscompare." It didn't. llvm-svn: 123215
* Support/Path: Deprecate PathV1::isDirectory and replace all uses with ↵Michael J. Spencer2011-01-113-3/+21
| | | | | | PathV2::is_directory. llvm-svn: 123209
* Turn on memdep's verifyRemoved() in an attempt to smoke out the cause of our ↵Jakob Stoklund Olesen2011-01-111-1/+3
| | | | | | gcc bootstrap miscompare. llvm-svn: 123207
* Teach constant folding to perform conversions from constant floatingChandler Carruth2011-01-112-55/+56
| | | | | | | | point values to their integer representation through the SSE intrinsic calls. This is the last part of a README.txt entry for which I have real world examples. llvm-svn: 123206
* Fix a random missed optimization by making InstCombine more aggressive when ↵Owen Anderson2011-01-112-17/+40
| | | | | | | | determining which bits are demanded by a comparison against a constant. llvm-svn: 123203
* Move ExpandAtomic into the integer expansion routines - it's only used there.Eric Christopher2011-01-112-86/+87
| | | | llvm-svn: 123202
* Even if we don't have 7 bytes of stack space we may need to save andEric Christopher2011-01-111-1/+8
| | | | | | | | restore the stack pointer from the frame pointer on thumbv6. Fixes rdar://8819685 llvm-svn: 123196
* Expand on the safeness of restoring the sp from the fp a bit more.Eric Christopher2011-01-101-1/+2
| | | | llvm-svn: 123193
* Fix PR 8916 (qv for analysis), at least the immediate problem.Dale Johannesen2011-01-101-1/+2
| | | | | | | | | | | There's an inherent tension in DAGCombine between assuming that things will be put in canonical form, and the Depth mechanism that disables transformations when recursion gets too deep. It would not surprise me if there's a lot of little bugs like this one waiting to be discovered. The mechanism seems fragile and I'd suggest looking at it from a design viewpoint. llvm-svn: 123191
* +0.0 vs -0.0 differences can be handled by looking at the user of theChris Lattner2011-01-101-3/+19
| | | | | | operation in some cases. llvm-svn: 123190
* McARM: Flush out hard coded known non-predicated mnemonic list.Daniel Dunbar2011-01-101-3/+12
| | | | llvm-svn: 123189
* McARM: Mark some T2 ...s instructions as codegen only, they aren't realDaniel Dunbar2011-01-101-4/+4
| | | | | | instructions but are restricted pseudo forms. llvm-svn: 123177
* ARM/MC: Mark several '...S' instructions as codegen only, they aren't realDaniel Dunbar2011-01-101-4/+6
| | | | | | instructions but are restricted pseudo forms. llvm-svn: 123176
* MC/ARM/AsmParser: Minor nitty fixes.Daniel Dunbar2011-01-101-3/+3
| | | | llvm-svn: 123175
* Fix merge falloutAnton Korobeynikov2011-01-101-2/+2
| | | | llvm-svn: 123172
* Update CMake stuffAnton Korobeynikov2011-01-1014-16/+15
| | | | llvm-svn: 123171
* Rename TargetFrameInfo into TargetFrameLowering. Also, put couple of FIXMEs ↵Anton Korobeynikov2011-01-10104-525/+546
| | | | | | and fixes here and there. llvm-svn: 123170
* MC/ARM/AsmParser: Split out SplitMnemonicAndCC().Daniel Dunbar2011-01-101-17/+29
| | | | llvm-svn: 123169
* Cleanup some of the constant folding code to consistently test intrinsicChandler Carruth2011-01-101-16/+18
| | | | | | | IDs when available rather than using a mixture of IDs and textual name comparisons. llvm-svn: 123165
* fit in 80 cols and use MBB::isSuccessor instead of a handChris Lattner2011-01-101-5/+4
| | | | | | rolled std::find. llvm-svn: 123164
* Teach instcombine about the rest of the SSE and SSE2 conversionChandler Carruth2011-01-101-4/+11
| | | | | | intrinsics element dependencies. Reviewed by Nick. llvm-svn: 123161
* Simplify a bunch of isVirtualRegister() and isPhysicalRegister() logic.Jakob Stoklund Olesen2011-01-1019-37/+33
| | | | | | | | These functions not longer assert when passed 0, but simply return false instead. No functional change intended. llvm-svn: 123155
* Fix Whitespace.Michael J. Spencer2011-01-102-17/+17
| | | | llvm-svn: 123152
* Support/Path: Deprecate PathV1::exists and replace all uses with ↵Michael J. Spencer2011-01-107-9/+21
| | | | | | PathV2::fs::exists. llvm-svn: 123151
* another random stab in the dark trying to fix llvm-gcc-i386-linux-selfhostChris Lattner2011-01-101-2/+4
| | | | llvm-svn: 123149
* another (more) aggressive attempt to bring llvm-gcc-i386-linux-selfhostChris Lattner2011-01-101-0/+2
| | | | | | back to life. llvm-svn: 123146
* expand on a noteChris Lattner2011-01-101-4/+9
| | | | llvm-svn: 123145
* temporarily disable memset formation from memsets in an effort to restore ↵Chris Lattner2011-01-091-0/+3
| | | | | | buildbot stability. llvm-svn: 123144
* typoChris Lattner2011-01-091-1/+1
| | | | llvm-svn: 123142
* xref a PR #Chris Lattner2011-01-091-22/+2
| | | | llvm-svn: 123141
* add a fixme: ir isn't expressive enough.Chris Lattner2011-01-091-0/+1
| | | | llvm-svn: 123139
* Step #4 in improving trip count analysis: HowFarToZero can analyzeChris Lattner2011-01-091-2/+11
| | | | | | | NUW AddRec's much more aggressively. We now get a trip count for @test2 in nsw.ll llvm-svn: 123138
* rearrange some code, no functionality change.Chris Lattner2011-01-091-41/+45
| | | | llvm-svn: 123136
* Add a note about the inability to model FP -> int conversions whichChandler Carruth2011-01-091-0/+55
| | | | | | | | perform rounding other than truncation in the IR. Common C code for this turns into really an LLVM intrinsic call that blocks a lot of further optimizations. llvm-svn: 123135
* Step #3 to improving trip count analysis: If we foldChris Lattner2011-01-091-4/+8
| | | | | | | | a + {b,+,stride} into {a+b,+,stride} (because a is LIV), then the resultant AddRec is NUW/NSW if the client says it is. llvm-svn: 123133
* Step #2 to improve trip count analysis for loops like this:Chris Lattner2011-01-091-6/+105
| | | | | | | | | | | | | | | | | | | void f(int* begin, int* end) { std::fill(begin, end, 0); } which turns into a != exit expression where one pointer is strided and (thanks to step #1) known to not overflow, and the other is loop invariant. The observation here is that, though the IV is strided by 4 in this case, that the IV *has* to become equal to the end value. It cannot "miss" the end value by stepping over it, because if it did, the strided IV expression would eventually wrap around. Handle this by turning A != B into "A-B != 0" where the A-B part is known to be NUW. llvm-svn: 123131
* Remove MachineRegisterInfo::getLastVirtReg(), it was giving wrong resultsJakob Stoklund Olesen2011-01-092-10/+9
| | | | | | | | | | | | | | | | | | when no virtual registers have been allocated. It was only used to resize IndexedMaps, so provide an IndexedMap::resize() method such that Map.grow(MRI.getLastVirtReg()); can be replaced with the simpler Map.resize(MRI.getNumVirtRegs()); This works correctly when no virtuals are allocated, and it bypasses the to/from index conversions. llvm-svn: 123130
* sort this.Chris Lattner2011-01-091-1/+1
| | | | llvm-svn: 123129
* Teach TargetRegisterInfo how to cram stack slot indexes in with the virtual andJakob Stoklund Olesen2011-01-096-14/+28
| | | | | | | | | | | | | physical register numbers. This makes the hack used in LiveInterval official, and lets LiveInterval be oblivious of stack slots. The isPhysicalRegister() and isVirtualRegister() predicates don't know about this, so when a variable may contain a stack slot, isStackSlot() should always be tested first. llvm-svn: 123128
* Add a note about a missed FP optimization.Chandler Carruth2011-01-091-0/+24
| | | | llvm-svn: 123126
* fix a few old bugs (found by inspection) where we would zap instructionsChris Lattner2011-01-091-1/+4
| | | | | | | | without informing memdep. This could cause nondeterminstic weirdness based on where instructions happen to get allocated, and will hopefully breath some life into some broken testers. llvm-svn: 123124
* Add a forgotten VireReg2IndexFunctor.Jakob Stoklund Olesen2011-01-091-1/+1
| | | | llvm-svn: 123123
OpenPOWER on IntegriCloud