summaryrefslogtreecommitdiffstats
path: root/llvm/lib/CodeGen/SelectionDAG
Commit message (Collapse)AuthorAgeFilesLines
* Fix two typos that Duncan spotted in a comment.Dan Gohman2009-01-291-1/+1
| | | | llvm-svn: 63312
* In the case of an extractelement on an insertelement value,Dan Gohman2009-01-291-1/+5
| | | | | | | the element indices may be equal if either one is not a constant. llvm-svn: 63311
* Revert r63273. This was already implemented by Dale. There's no need for myBill Wendling2009-01-291-72/+101
| | | | | | change. llvm-svn: 63301
* - Add DebugLoc to getTargetNode(). Bill Wendling2009-01-291-101/+72
| | | | | | | | | - Modify TableGen to add the DebugLoc when calling getTargetNode. (The light-weight wrappers are only temporary. The non-DebugLoc version will be removed once the whole debug info stuff is finished with.) llvm-svn: 63273
* Make x86's BT instruction matching more thorough, and add someDan Gohman2009-01-292-20/+73
| | | | | | | | | dagcombines that help it match in several more cases. Add several more cases to test/CodeGen/X86/bt.ll. This doesn't yet include matching for BT with an immediate operand, it just covers more register+register cases. llvm-svn: 63266
* Add DebugLoc-sensitive versions of many node creationDale Johannesen2009-01-291-0/+458
| | | | | | functions. Currently omitted: memcpy, memmove, memset. llvm-svn: 63259
* Add DebugLoc to the getNode() methods.Bill Wendling2009-01-281-41/+134
| | | | llvm-svn: 63245
* Add DebugLoc-aware constructors for SDNode derivedDale Johannesen2009-01-281-0/+23
| | | | | | | classes (those that reasonably have a DebugLoc associated with them). llvm-svn: 63236
* Fixed extract element when the result needs to be promoted and the input ↵Mon P Wang2009-01-281-0/+2
| | | | | | widened. llvm-svn: 63217
* Make isOperationLegal do what its name suggests, and introduce aDan Gohman2009-01-285-55/+62
| | | | | | | | | | | | | | | new isOperationLegalOrCustom, which does what isOperationLegal previously did. Update a bunch of callers to use isOperationLegalOrCustom instead of isOperationLegal. In some case it wasn't obvious which behavior is desired; when in doubt I changed then to isOperationLegalOrCustom as that preserves their previous behavior. This is for the second half of PR3376. llvm-svn: 63212
* Formatting.Duncan Sands2009-01-281-8/+5
| | | | llvm-svn: 63199
* Rename getAnalysisToUpdate to getAnalysisIfAvailable.Duncan Sands2009-01-281-2/+2
| | | | llvm-svn: 63198
* Use ValueType::bitsLT to simplify some code.Dan Gohman2009-01-282-2/+2
| | | | llvm-svn: 63170
* Use ZERO_EXTEND instead of ANY_EXTEND when promotingDan Gohman2009-01-282-2/+2
| | | | | | | shift amounts, to avoid implicitly assuming that target architectures will ignore the high bits. llvm-svn: 63169
* Add an assertion to the form of SelectionDAG::getConstant that takesDan Gohman2009-01-273-7/+13
| | | | | | | | a uint64_t to verify that the value is in range for the given type, to help catch accidental overflow. Fix a few places that relied on getConstant implicitly truncating the value. llvm-svn: 63128
* Delete redundant return statements.Dan Gohman2009-01-271-7/+0
| | | | llvm-svn: 63120
* Fix PR3393, which amounts to a bug in the expensiveDuncan Sands2009-01-262-16/+37
| | | | | | | | | checking logic. Rather than make the checking more complicated, I've tweaked some logic to make things conform to how the checking thought things ought to be, since this results in a simpler "mental model". llvm-svn: 63048
* During bittest switch lowering emit shift in the test block, which should ↵Anton Korobeynikov2009-01-261-11/+9
| | | | | | | | (theoretically) allow us to generate more efficient code. We don't do this now though :) llvm-svn: 63027
* Take the next steps in making SDUse more consistent with LLVM Use, andDan Gohman2009-01-263-82/+65
| | | | | | | | | | | | | | | | | tidy up SDUse and related code. - Replace the operator= member functions with a set method, like LLVM Use has, and variants setInitial and setNode, which take care up updating use lists, like LLVM Use's does. This simplifies code that calls these functions. - getSDValue() is renamed to get(), as in LLVM Use, though most places can either use the implicit conversion to SDValue or the convenience functions instead. - Fix some more node vs. value terminology issues. Also, eliminate the one remaining use of SDOperandPtr, and SDOperandPtr itself. llvm-svn: 62995
* Eliminate the loop that searches through each of the operandsDan Gohman2009-01-251-195/+178
| | | | | | | | | | | of each use in the SelectionDAG ReplaceAllUses* functions. Thanks to Chris for spotting this opportunity. Also, factor out code from all 5 of the ReplaceAllUses* functions into AddNonLeafNodeToCSEMaps, which is now renamed AddModifiedNodeToCSEMaps to more accurately reflect its purpose. llvm-svn: 62964
* Whitespace tidiments.Dan Gohman2009-01-251-2/+1
| | | | llvm-svn: 62963
* Move the N->use_empty() assert from DeleteNode toDan Gohman2009-01-251-3/+2
| | | | | | | DeleteNodeNotInCSEMaps, since DeleteNode just calls DeleteNodeNotInCSEMaps. llvm-svn: 62962
* Fix an indent and a typo.Nate Begeman2009-01-241-1/+1
| | | | llvm-svn: 62940
* Fold x-0 to x in unsafe-fp-math mode. This comes up in theDan Gohman2009-01-232-9/+19
| | | | | | | | | | | | | | testcase from PR3376, and in fact is sufficient to completely avoid the problem in that testcase. There's an underlying problem though; TLI.isOperationLegal considers Custom to be Legal, which might be ok in some cases, but that's what DAGCombiner is using in many places to test if something is legal when LegalOperations is true. When DAGCombiner is running after legalize, this isn't sufficient. I'll address this in a separate commit. llvm-svn: 62860
* Fix a minor bug in DAGCombiner's folding of SELECT. Folding "select C, 0, 1"Bob Wilson2009-01-221-3/+5
| | | | | | | | | | | | | | | | | | | | | | to "C ^ 1" is only valid when C is known to be either 0 or 1. Most of the similar foldings in this function only handle "i1" types, but this one appears intentionally written to handle larger integer types. If C has an integer type larger than "i1", this needs to check if the high bits of a boolean are known to be zero. I also changed the comment to describe this folding as "C ^ 1" instead of "~C", since that is what the code does and since the latter would only be valid for "i1" types. The good news is that most LLVM targets use TargetLowering::ZeroOrOneBooleanContent so this change will not disable the optimization; the bad news is that I've been unable to come up with a testcase to demonstrate the problem. I have also removed a "FIXME" comment for folding "select C, X, 0" to "C & X", since the code looks correct to me. It could be made more aggressive by not limiting the type to "i1", but that would then require checking for TargetLowering::ZeroOrNegativeOneBooleanContent. Similar changes could be done for the other SELECT foldings, but it was decided to be not worth the trouble and complexity (see e.g., r44663). llvm-svn: 62790
* Don't create ISD::FNEG nodes after legalize if they aren't legal.Dan Gohman2009-01-222-8/+30
| | | | | | | | | | | | | | Simplify x+0 to x in unsafe-fp-math mode. This avoids a bunch of redundant work in many cases, because in unsafe-fp-math mode, ISD::FADD with a constant is considered free to negate, so the DAGCombiner often negates x+0 to -0-x thinking it's free, when in reality the end result is -x, which is more expensive than x. Also, combine x*0 to 0. This fixes PR3374. llvm-svn: 62789
* Add SelectionDAG::getNOT method to construct bitwise NOT operations,Bob Wilson2009-01-224-26/+38
| | | | | | | corresponding to the "not" and "vnot" PatFrags. Use the new method in some places where it seems appropriate. llvm-svn: 62768
* Eliminate a couple of fields from TargetRegisterClass: SubRegClasses and ↵Evan Cheng2009-01-222-36/+5
| | | | | | SuperRegClasses. These are not necessary. Also eliminate getSubRegisterRegClass and getSuperRegisterRegClass. These are slow and their results can change if register file names change. Just use TargetLowering::getRegClassFor() to get the right TargetRegisterClass instead. llvm-svn: 62762
* fix a typoChris Lattner2009-01-221-1/+1
| | | | llvm-svn: 62761
* Simplify ReduceLoadWidth's logic: it doesn't need several differentDan Gohman2009-01-211-19/+10
| | | | | | | | special cases after producing the new reduced-width load, because the new load already has the needed adjustments built into it. This fixes several bugs due to the special cases, including PR3317. llvm-svn: 62692
* Cleanup whitespace and comments, and tweak someDuncan Sands2009-01-215-58/+53
| | | | | | | prototypes, in operand type legalization. No functionality change. llvm-svn: 62680
* CellSPU:Scott Michel2009-01-211-0/+4
| | | | | | | | | | | | | - Ensure that (operation) legalization emits proper FDIV libcall when needed. - Fix various bugs encountered during llvm-spu-gcc build, along with various cleanups. - Start supporting double precision comparisons for remaining libgcc2 build. Discovered interesting DAGCombiner feature, which is currently solved via custom lowering (64-bit constants are not legal on CellSPU, but DAGCombiner insists on inserting one anyway.) - Update README. llvm-svn: 62664
* Allow targets to legalize operations (with illegal operands) that produces ↵Sanjiv Gupta2009-01-215-68/+85
| | | | | | multiple values. For example, a load with an illegal operand (a load produces two values, a value and chain). llvm-svn: 62663
* Use "SINT_TO_FP" instead of "UINT_TO_FP" when getting the exponent. This wasBill Wendling2009-01-201-12/+13
| | | | | | | causing the limited precision stuff to produce the wrong result for values in the range [0, 1). llvm-svn: 62615
* Change TargetInstrInfo::isMoveInstr to return source and destination ↵Evan Cheng2009-01-201-2/+2
| | | | | | sub-register indices as well. llvm-svn: 62600
* Shift types need to match.Bill Wendling2009-01-201-8/+9
| | | | llvm-svn: 62571
* Fix a dagcombine to not generate loads of non-round integer types,Dan Gohman2009-01-201-1/+1
| | | | | | | as its comment says, even in the case where it will be generating extending loads. This fixes PR3216. llvm-svn: 62557
* Verify debug info.Devang Patel2009-01-192-9/+9
| | | | llvm-svn: 62545
* Remove SDNode's virtual destructor. This makes it impossible forDan Gohman2009-01-191-53/+14
| | | | | | | | | | | | | | SDNode subclasses to keep state that requires non-trivial destructors, however it was already effectively impossible, since the destructor isn't actually ever called. There currently aren't any SDNode subclasses affected by this, and in general it's desireable to keep SDNode objects light-weight. This eliminates the last virtual member function in the SDNode class, so it eliminates the need for a vtable pointer, making SDNode smaller. llvm-svn: 62539
* Fix SelectionDAG::ReplaceAllUsesWith to behave correctly whenDan Gohman2009-01-192-16/+39
| | | | | | | | | | | | | uses are added to the From node while it is processing From's use list, because of automatic local CSE. The fix is to avoid visiting any new uses. Fix a few places in the DAGCombiner that assumed that after a RAUW call, the From node has no users and may be deleted. This fixes PR3018. llvm-svn: 62533
* Few targets like PIC16 wants libcall generation for illegal type i16.Sanjiv Gupta2009-01-182-4/+16
| | | | llvm-svn: 62467
* Simplify extract element based on comments from Duncan Sands.Mon P Wang2009-01-181-7/+3
| | | | llvm-svn: 62459
* Simplify extract element of a scalar to vector.Mon P Wang2009-01-171-5/+13
| | | | llvm-svn: 62383
* Instead of adding dependence edges between terminator instructionsDan Gohman2009-01-162-9/+10
| | | | | | | | | | | | | | and every other instruction in their blocks to keep the terminator instructions at the end, teach the post-RA scheduler how to operate on ranges of instructions, and exclude terminators from the range of instructions that get scheduled. Also, exclude mid-block labels, such as EH_LABEL instructions, and schedule code before them separately from code after them. This fixes problems with the post-RA scheduler moving code past EH_LABELs. llvm-svn: 62366
* Use the getNode() accessor instead of accessing the NodeDan Gohman2009-01-161-1/+1
| | | | | | member directly, which is private as of r55504. llvm-svn: 62364
* new nodes should be added to the worklist, not old nodes.Chris Lattner2009-01-161-1/+2
| | | | llvm-svn: 62359
* CreateVirtualRegisters does trivial copy coalescing. If a node def is used ↵Evan Cheng2009-01-162-56/+63
| | | | | | by a single CopyToReg, it reuses the virtual register assigned to the CopyToReg. This won't work for SDNode that is a clone or is itself cloned. Disable this optimization for those nodes or it can end up with non-SSA machine instructions. llvm-svn: 62356
* Registry.h should not depend on CommandLine.h.Mikhail Glushenkov2009-01-161-0/+1
| | | | | | | Split Support/Registry.h into two files so that we have less to recompile every time CommandLine.h is changed. llvm-svn: 62312
* Delete trailing whitespace.Mikhail Glushenkov2009-01-161-234/+234
| | | | llvm-svn: 62307
* Initial hazard recognizer support in post-pass scheduling. This includesDan Gohman2009-01-161-1/+1
| | | | | | | a new toy hazard recognizier heuristic which attempts to direct the scheduler to avoid clumping large groups of loads or stores too densely. llvm-svn: 62291
OpenPOWER on IntegriCloud