| Commit message (Collapse) | Author | Age | Files | Lines | |
|---|---|---|---|---|---|
| * | restructure the top level non-local ptr dep query to handle | Chris Lattner | 2008-12-09 | 1 | -19/+26 |
| | | | | | | | | | the first block of a query specially. This makes the "complete query caching" subsystem more effective, avoiding predecessor queries. This speeds up GVN another 4%. llvm-svn: 60752 | ||||
| * | rename getNonLocalPointerDepInternal -> getNonLocalPointerDepFromBB | Chris Lattner | 2008-12-09 | 1 | -65/+83 |
| | | | | | | | | and split its inner loop out into a new GetNonLocalInfoForBlock function. No functionality change. llvm-svn: 60751 | ||||
| * | DisableGVCompilation should not abort on internal GlobalValue's. | Evan Cheng | 2008-12-09 | 1 | -2/+2 |
| | | | | | llvm-svn: 60750 | ||||
| * | In LegalizeOp, don't change the result type of CONVERT_RNDSAT when promoting | Mon P Wang | 2008-12-09 | 1 | -2/+2 |
| | | | | | | | one of its operand. llvm-svn: 60749 | ||||
| * | It's easy to handle SLE/SGE when the loop has a unit stride. | Nick Lewycky | 2008-12-09 | 1 | -1/+5 |
| | | | | | llvm-svn: 60748 | ||||
| * | if we have two elements, insert both, don't use std::sort. | Chris Lattner | 2008-12-09 | 1 | -3/+16 |
| | | | | | | | This speeds up the new GVN by another 3% llvm-svn: 60747 | ||||
| * | If we're only adding one new element to 'Cache', insert it into its known | Chris Lattner | 2008-12-09 | 1 | -1/+12 |
| | | | | | | | | position instead of using a full sort. This speeds up GVN by ~4% with the new memdep stuff. llvm-svn: 60746 | ||||
| * | convert a couple other places that use pred_iterator to use the caching | Chris Lattner | 2008-12-09 | 1 | -2/+4 |
| | | | | | | | pred iterator. llvm-svn: 60745 | ||||
| * | use hte new pred cache to speed up the new non-local memdep | Chris Lattner | 2008-12-09 | 1 | -4/+24 |
| | | | | | | | | queries. This speeds up GVN using the new queries (not yet checked in) by just over 10%. llvm-svn: 60743 | ||||
| * | Fix getNode to allow a vector for the shift amount for shifts of vectors. | Mon P Wang | 2008-12-09 | 2 | -9/+33 |
| | | | | | | | | | Fix the shift amount when unrolling a vector shift into scalar shifts. Fix problem in getShuffleScalarElt where it assumes that the input of a bit convert must be a vector. llvm-svn: 60740 | ||||
| * | Fix a really subtle off-by-one bug that Duncan noticed with valgrind | Chris Lattner | 2008-12-09 | 1 | -1/+1 |
| | | | | | | | on test/CodeGen/Generic/2007-06-06-CriticalEdgeLandingPad. llvm-svn: 60739 | ||||
| * | CellSPU: | Scott Michel | 2008-12-09 | 1 | -2/+5 |
| | | | | | | | | | - Change default scheduling preference to list-burr, which produces somewhat better code than the default. Could also use list-tdrr, but need to ask dev list about the appropriate handy mnemonic before commiting. llvm-svn: 60738 | ||||
| * | Add initial support for fast-isel of the [SU]ADDO intrinsics. It isn't | Bill Wendling | 2008-12-09 | 1 | -7/+84 |
| | | | | | | | | | | | | | | | | | | | | complete. For instance, it lowers the common case into this less-than-optimal code: addl %ecx, %eax seto %cl testb %cl, %cl jne LBB1_2 ## overflow instead of: addl %ecx, %eax jo LBB1_2 ## overflow That will come in a future commit. llvm-svn: 60737 | ||||
| * | Don't charge full latency for an anti-dependence, in this simplistic | Dan Gohman | 2008-12-09 | 1 | -3/+7 |
| | | | | | | | pipeline model. llvm-svn: 60733 | ||||
| * | Fix a couple of mistaken switch case fall-throughs. Thanks to Bill | Dan Gohman | 2008-12-08 | 1 | -0/+2 |
| | | | | | | | for spotting these! llvm-svn: 60728 | ||||
| * | remove DebugIterations option. Despite the accusations, | Chris Lattner | 2008-12-08 | 1 | -14/+1 |
| | | | | | | | | jump threading has been shown to only expose problems not have bugs itself. I'm sure it's completely bug free! ;-) llvm-svn: 60725 | ||||
| * | Re-apply 60689 now my head is screwed on right. | Evan Cheng | 2008-12-08 | 1 | -0/+2 |
| | | | | | llvm-svn: 60711 | ||||
| * | Fix the top-level comments, and fix some 80-column violations. | Dan Gohman | 2008-12-08 | 1 | -7/+9 |
| | | | | | llvm-svn: 60707 | ||||
| * | Revert 60689. It caused many regressions on Darwin targets. | Dan Gohman | 2008-12-08 | 1 | -3/+1 |
| | | | | | llvm-svn: 60705 | ||||
| * | Fix spelling. | Devang Patel | 2008-12-08 | 1 | -10/+10 |
| | | | | | | | Thanks Duncan! llvm-svn: 60702 | ||||
| * | Undo previous patch. | Devang Patel | 2008-12-08 | 1 | -3/+1 |
| | | | | | llvm-svn: 60701 | ||||
| * | Fix comment typo. | Duncan Sands | 2008-12-08 | 1 | -1/+1 |
| | | | | | llvm-svn: 60697 | ||||
| * | Factor out the code for sign-extending/truncating gep indices | Dan Gohman | 2008-12-08 | 2 | -26/+35 |
| | | | | | | | | | | and use it in x86 address mode folding. Also, make getRegForValue return 0 for illegal types even if it has a ValueMap for them, because Argument values are put in the ValueMap. This fixes PR3181. llvm-svn: 60696 | ||||
| * | add another level of caching for non-local pointer queries, keeping | Chris Lattner | 2008-12-08 | 1 | -7/+32 |
| | | | | | | | | | | track of whether the CachedNonLocalPointerInfo for a block is specific to a block. If so, just return it without any pred scanning. This is good for a 6% speedup on GVN (when it uses this lookup method, which it doesn't right now). llvm-svn: 60695 | ||||
| * | consistency | Chris Lattner | 2008-12-08 | 1 | -2/+2 |
| | | | | | llvm-svn: 60694 | ||||
| * | introduce a new RoundUpAlignment helper function, use it to | Chris Lattner | 2008-12-08 | 1 | -32/+29 |
| | | | | | | | | remove some more 64-bit divs and rems from the StructLayout ctor. llvm-svn: 60692 | ||||
| * | Make ConstantAggregateZero::get return a ConstantAggregateZero*, | Dan Gohman | 2008-12-08 | 1 | -1/+1 |
| | | | | | | | as suggested in PR3182. llvm-svn: 60691 | ||||
| * | Update CPP backend for PrintModulePass API changes. | Dan Gohman | 2008-12-08 | 1 | -4/+4 |
| | | | | | llvm-svn: 60690 | ||||
| * | Perform cheap checks first. | Evan Cheng | 2008-12-08 | 1 | -1/+3 |
| | | | | | llvm-svn: 60689 | ||||
| * | Some minor optimizations for isObjectSmallerThan. | Chris Lattner | 2008-12-08 | 1 | -7/+12 |
| | | | | | llvm-svn: 60687 | ||||
| * | Move createVirtualRegister out-of-line. | Dan Gohman | 2008-12-08 | 1 | -0/+18 |
| | | | | | llvm-svn: 60684 | ||||
| * | Clarify some comments. | Dan Gohman | 2008-12-08 | 1 | -2/+2 |
| | | | | | llvm-svn: 60683 | ||||
| * | Fixes for Visual Studio users. Patch by OvermindDL1 on llvm-dev! | Nick Lewycky | 2008-12-08 | 1 | -1/+1 |
| | | | | | llvm-svn: 60679 | ||||
| * | add an assert. the cast<> below would catch this but a message is more | Chris Lattner | 2008-12-07 | 1 | -0/+2 |
| | | | | | | | useful. llvm-svn: 60674 | ||||
| * | factor some code better. | Chris Lattner | 2008-12-07 | 1 | -17/+17 |
| | | | | | llvm-svn: 60673 | ||||
| * | factor some code, fixing some fixme's. | Chris Lattner | 2008-12-07 | 1 | -32/+23 |
| | | | | | llvm-svn: 60672 | ||||
| * | add support for caching pointer dependence queries. Nothing uses this yet | Chris Lattner | 2008-12-07 | 1 | -17/+211 |
| | | | | | | | so it "can't" break anything. That said, it does appear to work. llvm-svn: 60654 | ||||
| * | fix a bug I introduced in simplifycfg handling single entry phi | Chris Lattner | 2008-12-07 | 1 | -1/+0 |
| | | | | | | | | nodes. FoldSingleEntryPHINodes deletes the PHI, so there is no need to delete it afterward. llvm-svn: 60653 | ||||
| * | Switch to top-down mode and fix a crasher this exposed caused by an error in the | Owen Anderson | 2008-12-07 | 1 | -1/+11 |
| | | | | | | | live interval updating. llvm-svn: 60652 | ||||
| * | Fix typo, psuedo -> pseudo. | Nick Lewycky | 2008-12-07 | 2 | -35/+35 |
| | | | | | llvm-svn: 60651 | ||||
| * | Some internal refactoring to make it easier to cache results. | Chris Lattner | 2008-12-07 | 1 | -17/+26 |
| | | | | | llvm-svn: 60650 | ||||
| * | Introduce a new MemDep::getNonLocalPointerDependency | Chris Lattner | 2008-12-07 | 1 | -13/+82 |
| | | | | | | | | | | method. This will eventually take over load/store dep queries from getNonLocalDependency. For now it works fine, but is incredibly slow because it does no caching. Lets not switch GVN to use it until that is fixed :) llvm-svn: 60649 | ||||
| * | push the "pointer case" up the analysis stack a bit. This causes | Chris Lattner | 2008-12-07 | 1 | -47/+83 |
| | | | | | | | | duplication of logic (in 2 places) to determine what pointer a load/store touches. This will be addressed in a future commit. llvm-svn: 60648 | ||||
| * | make clients have to know how to call getCallSiteDependencyFrom | Chris Lattner | 2008-12-07 | 1 | -2/+13 |
| | | | | | | | instead of making getDependencyFrom do it. llvm-svn: 60647 | ||||
| * | rename some variables for consistency | Chris Lattner | 2008-12-07 | 1 | -6/+6 |
| | | | | | llvm-svn: 60644 | ||||
| * | I love how using out of scope variables is not an error with GCC, no really ↵ | Chris Lattner | 2008-12-07 | 1 | -3/+3 |
| | | | | | | | I do. llvm-svn: 60643 | ||||
| * | Rename getCallSiteDependency -> getCallSiteDependencyFrom to | Chris Lattner | 2008-12-07 | 1 | -4/+5 |
| | | | | | | | | | emphasize the scanning and make it more similar to getDependencyFrom llvm-svn: 60642 | ||||
| * | a memdep query on a volatile load/store will always return | Chris Lattner | 2008-12-07 | 1 | -11/+10 |
| | | | | | | | | | | clobber with the current implementation. Instead of returning a "precise clobber" just return a fuzzy one. This doesn't matter to any clients anyway and should speed up analysis time very very slightly. llvm-svn: 60641 | ||||
| * | don't bother touching volatile stores, they will just return clobber on | Chris Lattner | 2008-12-07 | 1 | -0/+5 |
| | | | | | | | everything interesting anyway. llvm-svn: 60640 | ||||
| * | remove the ability to get memdep info for vaarg. I don't think the | Chris Lattner | 2008-12-07 | 1 | -6/+9 |
| | | | | | | | original impl was correct and noone actually makes the query anyway. llvm-svn: 60639 | ||||

