summaryrefslogtreecommitdiffstats
path: root/llvm/lib
Commit message (Collapse)AuthorAgeFilesLines
* Tidy up BasicBlock::getFirstNonPHI, and change a bunch of places toDan Gohman2008-05-2315-61/+34
| | | | | | use it instead of duplicating its functionality. llvm-svn: 51499
* Add #includes to make some dependencies explicit.Dan Gohman2008-05-234-0/+4
| | | | llvm-svn: 51496
* Issue errors in several situations instead of aborting.Dan Gohman2008-05-231-7/+21
| | | | llvm-svn: 51493
* Elaborate on the entry on integer vector multiplication by constants.Dan Gohman2008-05-231-1/+6
| | | | llvm-svn: 51491
* Fix a duplicated pattern.Evan Cheng2008-05-231-2/+2
| | | | llvm-svn: 51490
* Use PMULDQ for v2i64 multiplies when SSE4.1 is available. And addDan Gohman2008-05-233-7/+13
| | | | | | load-folding table entries for PMULDQ and PMULLD. llvm-svn: 51489
* New entry.Evan Cheng2008-05-231-0/+44
| | | | llvm-svn: 51487
* Rewrite a loop to avoid using iterators pointing toDale Johannesen2008-05-231-4/+6
| | | | | | | elements that have been erased. Based on a patch by Nicolas Capens. llvm-svn: 51485
* Fix another isFirstClassType that now needs to be isSingleValueType.Dan Gohman2008-05-231-1/+1
| | | | | | This fixes recent CBE regressions. llvm-svn: 51483
* Replace some weird usage of UserOp1 introduced in r49492 by a plain if.Matthijs Kooijman2008-05-231-1/+3
| | | | llvm-svn: 51482
* Restucture a part of the SimplifyCFG pass and include a testcase.Matthijs Kooijman2008-05-231-65/+99
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The SimplifyCFG pass looks at basic blocks that contain only phi nodes, followed by an unconditional branch. In a lot of cases, such a block (BB) can be merged into their successor (Succ). This merging is performed by TryToSimplifyUncondBranchFromEmptyBlock. It does this by taking all phi nodes in the succesor block Succ and expanding them to include the predecessors of BB. Furthermore, any phi nodes in BB are moved to Succ and expanded to include the predecessors of Succ as well. Before attempting this merge, CanPropagatePredecessorsForPHIs checks to see if all phi nodes can be properly merged. All functional changes are made to this function, only comments were updated in TryToSimplifyUncondBranchFromEmptyBlock. In the original code, CanPropagatePredecessorsForPHIs looks quite convoluted and more like stack of checks added to handle different kinds of situations than a comprehensive check. In particular the first check in the function did some value checking for the case that BB and Succ have a common predecessor, while the last check in the function simply rejected all cases where BB and Succ have a common predecessor. The first check was still useful in the case that BB did not contain any phi nodes at all, though, so it was not completely useless. Now, CanPropagatePredecessorsForPHIs is restructured to to look a lot more similar to the code that actually performs the merge. Both functions now look at the same phi nodes in about the same order. Any conflicts (phi nodes with different values for the same source) that could arise from merging or moving phi nodes are detected. If no conflicts are found, the merge can happen. Apart from only restructuring the checks, two main changes in functionality happened. Firstly, the old code rejected blocks with common predecessors in most cases. The new code performs some extra checks so common predecessors can be handled in a lot of cases. Wherever common predecessors still pose problems, the blocks are left untouched. Secondly, the old code rejected the merge when values (phi nodes) from BB were used in any other place than Succ. However, it does not seem that there is any situation that would require this check. Even more, this can be proven. Consider that BB is a block containing of a single phi node "%a" and a branch to Succ. Now, since the definition of %a will dominate all of its uses, BB will dominate all blocks that use %a. Furthermore, since the branch from BB to Succ is unconditional, Succ will also dominate all uses of %a. Now, assume that one predecessor of Succ is not dominated by BB (and thus not dominated by Succ). Since at least one use of %a (but in reality all of them) is reachable from Succ, you could end up at a use of %a without passing through it's definition in BB (by coming from X through Succ). This is a contradiction, meaning that our original assumption is wrong. Thus, all predecessors of Succ must also be dominated by BB (and thus also by Succ). This means that moving the phi node %a from BB to Succ does not pose any problems when the two blocks are merged, and any use checks are not needed. llvm-svn: 51478
* Indent fix.Matthijs Kooijman2008-05-231-2/+2
| | | | llvm-svn: 51477
* Constant integer vectors may also be negated.Nick Lewycky2008-05-231-0/+5
| | | | llvm-svn: 51476
* Typo.Nick Lewycky2008-05-231-1/+1
| | | | llvm-svn: 51475
* Revert X + X --> X * 2 optz'n which pessimizes heavily on x86.Nick Lewycky2008-05-231-13/+5
| | | | llvm-svn: 51474
* we compile multiply-by-constant into horrible code. Doesn't sse4 have someChris Lattner2008-05-231-0/+38
| | | | | | instruction for doing this? llvm-svn: 51473
* Implement X + X for vectors.Nick Lewycky2008-05-231-5/+13
| | | | llvm-svn: 51472
* Fix a recently added optimization to not crash on vectors.Nick Lewycky2008-05-231-2/+10
| | | | llvm-svn: 51471
* Generalize the new code in instcombine's ComputeNumSignBits for handlingDan Gohman2008-05-232-51/+22
| | | | | | | and/or to handle more cases (such as this add-sitofp.ll testcase), and port it to selectiondag's ComputeNumSignBits. llvm-svn: 51469
* Make structs and arrays first-class types, and add assemblyDan Gohman2008-05-238-1261/+1633
| | | | | | | | | and bitcode support for the extractvalue and insertvalue instructions and constant expressions. Note that this does not yet include CodeGen support. llvm-svn: 51468
* Use isSingleValueType instead of isFirstClassType toDan Gohman2008-05-231-2/+2
| | | | | | exclude struct and array types. llvm-svn: 51467
* Remove warnings about comparison between signed and unsigned expressions.Bill Wendling2008-05-231-2/+2
| | | | llvm-svn: 51465
* Allow for switch with no cases. Was causing faultDale Johannesen2008-05-231-0/+2
| | | | | | in gcc.dg/pr27531-1.c. llvm-svn: 51464
* Bug: rcpps can only folds a load if the address is 16-byte aligned. Fixed ↵Evan Cheng2008-05-231-57/+79
| | | | | | | | many 'ps' load folding patterns in X86InstrSSE.td which are missing the proper alignment checks. Also fixed some 80 col. violations. llvm-svn: 51462
* Add more IR support for the new extractvalue and insertvalueDan Gohman2008-05-232-12/+79
| | | | | | instructions. llvm-svn: 51461
* Use isSingleValueType instead of isFirstClassType toDan Gohman2008-05-231-2/+2
| | | | | | exclude struct and array types. llvm-svn: 51460
* Use isSingleValueType instead of isFirstClassType toDan Gohman2008-05-233-6/+6
| | | | | | exclude struct and array types. llvm-svn: 51459
* Put const weak stuff in appropriate section on Darwin.Dale Johannesen2008-05-231-1/+9
| | | | | | g++.dg/abi/key2.C llvm-svn: 51458
* Use isSingleValueType instead of isFirstClassType toDan Gohman2008-05-232-2/+2
| | | | | | exclude struct and array types. llvm-svn: 51456
* X86CodeEmitter should not set PIC style to None at initialization time. This ↵Evan Cheng2008-05-221-3/+1
| | | | | | will break codegen if relocation model is changed to PIC_ later. llvm-svn: 51455
* Use isSingleValueType instead of isFirstClassType toDan Gohman2008-05-221-1/+1
| | | | | | exclude struct and array types. llvm-svn: 51452
* Recognize the "default" keyword, which is documented in LangRef.htmlDan Gohman2008-05-221-0/+1
| | | | | | and supported in the grammar, in the lexer. llvm-svn: 51448
* When rewriting defs and uses after spilling, don't set the weight of aDavid Greene2008-05-221-1/+3
| | | | | | | | live interval to infinity if the instruction being rewritten is an original remat def instruction. We were only checking against the clone of the remat def which doesn't actually appear in the IR at all. llvm-svn: 51440
* Don't attempt to update SpillSlotToUsesMap for stack slots that aren'tDavid Greene2008-05-221-2/+14
| | | | | | generated by the spiller. llvm-svn: 51439
* transform more loops to iterator form, detabifyGabor Greif2008-05-222-7/+8
| | | | llvm-svn: 51436
* Add missing patterns.Evan Cheng2008-05-221-0/+10
| | | | llvm-svn: 51435
* Rewrite operand loops to use iterators. This shrinks .o file (at gcc4.0.1 ↵Gabor Greif2008-05-221-4/+5
| | | | | | -O3 x86) substantially (>500 bytes). Reason still unknown. llvm-svn: 51423
* fix an off-by-one error in my previous patch, don't treat the callee as a ↵Chris Lattner2008-05-221-10/+9
| | | | | | incoming arg. llvm-svn: 51422
* Add support for multiple-return values in inline asm. This shouldChris Lattner2008-05-221-34/+42
| | | | | | | get inline asm working as well as it did previously with the CBE with the new MRV support for inline asm. llvm-svn: 51420
* Fix PR2267, by allowing indirect outputs to be intermixedChris Lattner2008-05-221-1/+4
| | | | | | | with normal outputs. Testcase here: test/CodeGen/X86/asm-indirect-mem.ll llvm-svn: 51409
* Free and vaarg are not really volatile.Chris Lattner2008-05-221-9/+2
| | | | llvm-svn: 51407
* rewrite the validity checking for memory promotion to be simpler,Chris Lattner2008-05-221-49/+52
| | | | | | | more aggressive, and more correct. Verify that we only attempt to promote loads and stores. llvm-svn: 51406
* Use 'continue' to reduce nesting in this loop. No functionality change.Chris Lattner2008-05-221-54/+51
| | | | llvm-svn: 51399
* Fix PR2343. An *interesting* coalescer bug.Evan Cheng2008-05-211-1/+27
| | | | | | | | | | | | | | BB1: vr1025 = copy vr1024 .. BB2: vr1024 = op = op vr1025 <loop eventually branch back to BB1> Even though vr1025 is copied from vr1024, it's not safe to coalesced them since live range of vr1025 intersects the def of vr1024. This happens when vr1025 is assigned the value of the previous iteration of vr1024 in the loop. llvm-svn: 51394
* Follow-up to the reverting of r51218. This puts the checks out-of-line. BecauseBill Wendling2008-05-211-0/+12
| | | | | | | | they aren't in the header file, systems with a <string> header file that isn't 64-bit clean shouldn't warn if #including Path.h and specifying -Wshorten-64-to-32. llvm-svn: 51393
* Fix a couple issues with the JIT and multiple modules:Nate Begeman2008-05-213-8/+75
| | | | | | | | | | | | 1. The "JITState" object creates a PassManager with the ModuleProvider that the jit is created with. If the ModuleProvider is removed and deleted, the PassManager is invalid. 2. The Global maps in the JIT were not invalidated with a ModuleProvider was removed. This could lead to a case where the Module would be freed, and a new Module with Globals at the same addresses could return invalid results. llvm-svn: 51384
* suppress gcc3.4.6's <no value returned> warningsGabor Greif2008-05-211-0/+2
| | | | llvm-svn: 51372
* When LSR is replacing an instruction, callDan Gohman2008-05-211-7/+10
| | | | | | | | | | | | | | | | | | | | ScalarEvolution::deleteValueFromRecords on it before doing the replaceAllUsesWith, because ScalarEvolution looks at the instruction's users to find SCEV references to the instruction's SCEV object in its internal maps. Move all of LSR's loop-related state clearing after processing the loop and before cleaning up dead PHI nodes. This eliminates all of LSR's SCEV references just before the calls to ScalarEvolution::deleteValueFromRecords so that when ScalarEvolution drops its own SCEV references, the reference counts will reach zero and the SCEVs will be deleted immediately. These changes fix some compiler aborts involving ScalarEvolution holding onto and reusing SCEV objects for instructions that have been deleted. No regression test unfortunately; because the symptoms were due to dangling pointers, reduced testcases ended up being fairly arbitrary. llvm-svn: 51359
* Fix PR2346 by marking vaarg as volatile so that licm doesn't try toChris Lattner2008-05-201-1/+4
| | | | | | hoist them. llvm-svn: 51356
* Port SelectionDAG's ComputeNumSignBits-using code to instcombine,Dan Gohman2008-05-201-1/+28
| | | | | | now that instcombine also has ComputeNumSignBits. llvm-svn: 51350
OpenPOWER on IntegriCloud