summaryrefslogtreecommitdiffstats
path: root/llvm/lib/CodeGen
Commit message (Collapse)AuthorAgeFilesLines
* Handle spilling around an instruction that has an early-clobber re-definition ofJakob Stoklund Olesen2011-04-181-14/+15
| | | | | | | | the spilled register. This is quite common on ARM now that some stores have early-clobber defines. llvm-svn: 129714
* Fix a bug where we were counting the alias sets as completely usedEric Christopher2011-04-181-5/+0
| | | | | | | | | registers for fast allocation a different way. This has us updating used registers only when we're using that exact register. Fixes rdar://9207598 llvm-svn: 129711
* while we're at it, handle 'sdiv exact' of a power of 2 also,Chris Lattner2011-04-181-0/+8
| | | | | | this fixes a few rejects on c++ iterator loops. llvm-svn: 129694
* fix rdar://9297011 - udiv by power of two causing fast-isel rejectsChris Lattner2011-04-181-0/+4
| | | | llvm-svn: 129693
* 1. merge fast-isel-shift-imm.ll into fast-isel-x86-64.llChris Lattner2011-04-171-18/+30
| | | | | | | | | | 2. implement rdar://9289501 - fast isel should fold trivial multiplies to shifts 3. teach tblgen to handle shift immediates that are different sizes than the shifted operands, eliminating some code from the X86 fast isel backend. 4. Have FastISel::SelectBinaryOp use (the poorly named) FastEmit_ri_ function instead of FastEmit_ri to simplify code. llvm-svn: 129666
* fix an oversight which caused us to compile the testcase (and otherChris Lattner2011-04-171-5/+3
| | | | | | | | | | | | | | | | | | | less trivial things) into a dummy lea. Before we generated: _test: ## @test movq _G@GOTPCREL(%rip), %rax leaq (%rax), %rax ret now we produce: _test: ## @test movq _G@GOTPCREL(%rip), %rax ret This is part of rdar://9289558 llvm-svn: 129662
* Fix rdar://9289512 - not folding load into compare at -O0Chris Lattner2011-04-171-15/+41
| | | | | | | | | | | | | | | | | | | | | | | | | | | The basic issue here is that bottom-up isel is matching the branch and compare, and was failing to fold the load into the branch/compare combo. Fixing this (by allowing folding into any instruction of a sequence that is selected) allows us to produce things like: cmpb $0, 52(%rax) je LBB4_2 instead of: movb 52(%rax), %cl cmpb $0, %cl je LBB4_2 This makes the generated -O0 code run a bit faster, but also speeds up compile time by putting less pressure on the register allocator and generating less code. This was one of the biggest classes of missing load folding. Implementing this shrinks 176.gcc's c-decl.s (as a random example) by about 4% in (verbose-asm) line count. llvm-svn: 129656
* split a complex predicate out to a helper function. Simplify two for loops,Chris Lattner2011-04-171-10/+16
| | | | | | | which don't need to check for falling off the end of a block *and* end of phi nodes, since terminators are never phis. llvm-svn: 129655
* fix rdar://9289583 - fast isel should handle non-canonical commutative binopsChris Lattner2011-04-171-4/+23
| | | | | | | | | | allowing us to fold the immediate into the 'and' in this case: int test1(int i) { return 8&i; } llvm-svn: 129653
* PR9055: extend the fix to PR4050 (r70179) to apply to zext and anyext.Eli Friedman2011-04-161-2/+2
| | | | | | | Returning a new node makes the code try to replace the old node, which in the included testcase is killed by CSE. llvm-svn: 129650
* Unbreak the MSVC 2010 build.Francois Pichet2011-04-161-1/+2
| | | | | | For further information on this particular issue see: http://connect.microsoft.com/VisualStudio/feedback/details/520043/error-converting-from-null-to-a-pointer-type-in-std-pair llvm-svn: 129642
* Remove unused variable.Benjamin Kramer2011-04-161-5/+2
| | | | llvm-svn: 129639
* Put each personality function in a section. This fixes the gnu ld warning:Rafael Espindola2011-04-162-8/+36
| | | | | | error in foo.o; no .eh_frame_hdr table will be created. llvm-svn: 129635
* Fix divmod libcall lowering. Convert to {S|U}DIVREM first and then expand ↵Evan Cheng2011-04-161-71/+65
| | | | | | the node to a libcall. rdar://9280991 llvm-svn: 129633
* Introduce support to encode Objective-C property information in debugging ↵Devang Patel2011-04-161-0/+31
| | | | | | information generated for an interface. llvm-svn: 129624
* Some refactoring suggested by Anton Korobeynikov.Rafael Espindola2011-04-152-41/+29
| | | | llvm-svn: 129600
* Teach the SplitKit blitter to handle multiply defined values as well.Jakob Stoklund Olesen2011-04-152-96/+223
| | | | | | | | | | | | | | The transferValues() function can now handle both singly and multiply defined values, as long as the resulting live range is known. Only rematerialized values have their live range recomputed by extendRange(). The updateSSA() function can now insert PHI values in bulk across multiple values in multiple target registers in one pass. The list of blocks received from transferValues() is in layout order which seems to work well for the iterative algorithm. Blocks from extendRange() are still in reverse BFS order, but this function is used so rarely now that it doesn't matter. llvm-svn: 129580
* Remember to set flag.Jakob Stoklund Olesen2011-04-151-0/+1
| | | | llvm-svn: 129579
* Add 129518 back with a fix for when we are producing eh just because of ↵Rafael Espindola2011-04-152-16/+71
| | | | | | | | | debug info. Change ELF systems to use CFI for producing the EH tables. This reduces the size of the clang binary in Debug builds from 690MB to 679MB. llvm-svn: 129571
* Fix a ton of comment typos found by codespell. Patch byChris Lattner2011-04-1522-32/+31
| | | | | | Luis Felipe Strano Moraes! llvm-svn: 129558
* Revert r129518, "Change ELF systems to use CFI for producing the EH tables. ↵NAKAMURA Takumi2011-04-152-70/+15
| | | | | | | | This reduces the" It broke several builds. llvm-svn: 129557
* Fix another instance of the DAG combiner not using the correct type for the ↵Owen Anderson2011-04-141-3/+5
| | | | | | RHS of a shift. llvm-svn: 129522
* Change ELF systems to use CFI for producing the EH tables. This reduces theRafael Espindola2011-04-142-15/+70
| | | | | | size of the clang binary in Debug builds from 690MB to 679MB. llvm-svn: 129518
* In the pre-RA scheduler, maintain cmp+br proximity.Andrew Trick2011-04-142-13/+61
| | | | | | | | | | | | | | | | | | | | | | | | This is done by pushing physical register definitions close to their use, which happens to handle flag definitions if they're not glued to the branch. This seems to be generally a good thing though, so I didn't need to add a target hook yet. The primary motivation is to generate code closer to what people expect and rule out missed opportunity from enabling macro-op fusion. As a side benefit, we get several 2-5% gains on x86 benchmarks. There is one regression: SingleSource/Benchmarks/Shootout/lists slows down be -10%. But this is an independent scheduler bug that will be tracked separately. See rdar://problem/9283108. Incidentally, pre-RA scheduling is only half the solution. Fixing the later passes is tracked by: <rdar://problem/8932804> [pre-RA-sched] on x86, attempt to schedule CMP/TEST adjacent with condition jump Fixes: <rdar://problem/9262453> Scheduler unnecessary break of cmp/jump fusion llvm-svn: 129508
* sink a call into its only use.Chris Lattner2011-04-141-2/+1
| | | | llvm-svn: 129503
* During post-legalization DAG combining, be careful to only create shifts ↵Owen Anderson2011-04-131-1/+8
| | | | | | where the RHS is of the legal type for the new operation. llvm-svn: 129484
* Remove extra bytes that were added for gdb. We do not have good poiner to ↵Devang Patel2011-04-131-8/+1
| | | | | | understand actual reason behind this fixme. Spot checking suggest that newer gdb does not need this. llvm-svn: 129461
* Stop using dead function.Jakob Stoklund Olesen2011-04-133-18/+0
| | | | llvm-svn: 129442
* Recommit r129383. PreRA scheduler heuristic fixes: VRegCycle, TokenFactor ↵Andrew Trick2011-04-132-156/+190
| | | | | | | | | | | | | | | | | | | | | latency. Additional fixes: Do something reasonable for subtargets with generic itineraries by handle node latency the same as for an empty itinerary. Now nodes default to unit latency unless an itinerary explicitly specifies a zero cycle stage or it is a TokenFactor chain. Original fixes: UnitsSharePred was a source of randomness in the scheduler: node priority depended on the queue data structure. I rewrote the recent VRegCycle heuristics to completely replace the old heuristic without any randomness. To make the ndoe latency adjustments work, I also needed to do something a little more reasonable with TokenFactor. I gave it zero latency to its consumers and always schedule it as low as possible. llvm-svn: 129421
* Temporarily revert r129408 to see if it brings the bots back.Eric Christopher2011-04-131-0/+2
| | | | llvm-svn: 129417
* Fix a bug where we were counting the alias sets as completely usedEric Christopher2011-04-121-2/+0
| | | | | | | | registers for fast allocation. Fixes rdar://9207598 llvm-svn: 129408
* I missed this new file in previous commit.Devang Patel2011-04-121-0/+973
| | | | llvm-svn: 129407
* Simplify. There is no need to use static variable.Devang Patel2011-04-121-3/+1
| | | | llvm-svn: 129406
* Do not reuse parameter name.Devang Patel2011-04-121-1/+1
| | | | llvm-svn: 129405
* This mechanical patch moves type handling into CompileUnit from DwarfDebug. ↵Devang Patel2011-04-124-1304/+377
| | | | | | In case of multiple compile unit in one object file, each compile unit is responsible for its own set of type entries anyway. This refactoring makes this obvious. llvm-svn: 129402
* Add more comments... err debug statements to the fast allocator.Eric Christopher2011-04-121-3/+16
| | | | llvm-svn: 129400
* SparseBitVector is SLOW.Jakob Stoklund Olesen2011-04-122-48/+58
| | | | | | | Use a Bitvector instead, we didn't need the smaller memory footprint anyway. This makes the greedy register allocator 10% faster. llvm-svn: 129390
* Revert 129383. It causes some targets to hit a scheduler assert.Andrew Trick2011-04-122-184/+157
| | | | llvm-svn: 129385
* PreRA scheduler heuristic fixes: VRegCycle, TokenFactor latency.Andrew Trick2011-04-122-157/+184
| | | | | | | | | | | | UnitsSharePred was a source of randomness in the scheduler: node priority depended on the queue data structure. I rewrote the recent VRegCycle heuristics to completely replace the old heuristic without any randomness. To make these heuristic adjustments to node latency work, I also needed to do something a little more reasonable with TokenFactor. I gave it zero latency to its consumers and always schedule it as low as possible. llvm-svn: 129383
* Create new intervals for isolated blocks during region splitting.Jakob Stoklund Olesen2011-04-123-37/+46
| | | | | | | | | This merges the behavior of splitSingleBlocks into splitAroundRegion, so the RS_Region and RS_Block register stages can be coalesced. That means the leftover intervals after region splitting go directly to spilling instead of a second pass of per-block splitting. llvm-svn: 129379
* Add SplitKit API to query and select the current interval being worked on.Jakob Stoklund Olesen2011-04-122-2/+17
| | | | | | This makes it possible to target multiple registers in one pass. llvm-svn: 129374
* Fix a bug in RegAllocBase::addMBBLiveIns() where a basic block could ↵Jakob Stoklund Olesen2011-04-121-1/+1
| | | | | | accidentally be skipped. llvm-svn: 129373
* Remove dead typedef.Devang Patel2011-04-121-2/+0
| | | | llvm-svn: 129368
* Refactor CompileUnit into a separate header.Devang Patel2011-04-122-98/+124
| | | | llvm-svn: 129367
* Fix typo.Eric Christopher2011-04-121-1/+1
| | | | llvm-svn: 129334
* Reuse live interval union between functions. This saves a bit of compile timeJakob Stoklund Olesen2011-04-112-4/+11
| | | | | | when compiling many small functions. llvm-svn: 129321
* Just because a GlobalVariable's initializer is [N x { i32, void ()* }] doesn'tNick Lewycky2011-04-111-1/+5
| | | | | | | | | | | | mean that it has to be ConstantArray of ConstantStruct. We might have ConstantAggregateZero, at either level, so don't crash on that. Also, semi-deprecate the sentinal value. The linker isn't aware of sentinals so we end up with the two lists appended, each with their "sentinals" on them. Different parts of LLVM treated sentinals differently, so make them all just ignore the single entry and continue on with the rest of the list. llvm-svn: 129307
* Speed up eviction by stopping collectInterferingVRegs as soon as the spillJakob Stoklund Olesen2011-04-113-14/+23
| | | | | | weight limit has been exceeded. llvm-svn: 129305
* The default of the dispatch switch statement was to branch to a BB that executedBill Wendling2011-04-111-7/+8
| | | | | | | | | | | | | | | | | | | | the 'unwind' instruction. However, later on that instruction was converted into a jump to the basic block it was located in, causing an infinite loop when we get there. It turns out, we get there if the _Unwind_Resume_or_Rethrow call returns (which it's not supposed to do). It returns if it cannot find a place to unwind to. Thus we would get what appears to be a "hang" when in reality it's just that the EH couldn't be propagated further along. Instead of infinitely looping (or calling `unwind', which none of our back-ends support (it's lowered into nothing...)), call the @llvm.trap() intrinsic instead. This may not conform to specific rules of a particular language, but it's rather better than infinitely looping. <rdar://problem/9175843&9233582> llvm-svn: 129302
* Look pass copies when determining whether hoisting would end up inserting ↵Evan Cheng2011-04-111-8/+17
| | | | | | more copies. rdar://9266679 llvm-svn: 129297
OpenPOWER on IntegriCloud