summaryrefslogtreecommitdiffstats
path: root/llvm/lib/CodeGen
Commit message (Collapse)AuthorAgeFilesLines
...
* Add/complete support for integer and floatDuncan Sands2008-06-253-38/+282
| | | | | | | | select_cc and friends. This code could be factorized a bit but I'm not sure that it's worth it. llvm-svn: 52724
* Remove the OrigVT member from AtomicSDNode, as it is redundant withDan Gohman2008-06-253-13/+9
| | | | | | the base SDNode's VTList. llvm-svn: 52722
* Added MemOperands to Atomic operations since Atomics touches memory.Mon P Wang2008-06-253-34/+93
| | | | | | | | Added abstract class MemSDNode for any Node that have an associated MemOperand Changed atomic.lcs => atomic.cmp.swap, atomic.las => atomic.load.add, and atomic.lss => atomic.load.sub llvm-svn: 52706
* Enable two-address remat by default.Evan Cheng2008-06-251-35/+33
| | | | llvm-svn: 52701
* Use SmallVector instead of std::vector for a minor compile time improvement.Owen Anderson2008-06-241-11/+11
| | | | llvm-svn: 52689
* A brief survey of priority_queue usage in the tree turned this upDan Gohman2008-06-231-1/+1
| | | | | | as a questionable case, but the code isn't actually needed. llvm-svn: 52657
* This situation can occur:Bill Wendling2008-06-231-2/+17
| | | | | | | | | | | | | | | | | | | ,------. | | | v | t2 = phi ... t1 ... | | | v | t1 = ... | ... = ... t1 ... | | `------' where there is a use in a PHI node that's a predecessor to the defining block. We don't want to mark all predecessors as having the value "alive" in this case. Also, the assert was too restrictive and didn't handle this case. llvm-svn: 52655
* Use the new PriorityQueue in ScheduleDAGList too, which alsoDan Gohman2008-06-231-19/+3
| | | | | | needs arbitrary-element removal. llvm-svn: 52654
* Use getMBBEndIdx rather than assuming that the end is right after the last ↵Owen Anderson2008-06-231-1/+1
| | | | | | instruction in the block. llvm-svn: 52649
* Remove option used to debug stack coloring bugs. It's no longer needed since ↵Evan Cheng2008-06-231-6/+1
| | | | | | stack coloring is now bug free. llvm-svn: 52644
* Move a DenseMap's declaration outside of a loop, and just callDan Gohman2008-06-231-1/+2
| | | | | | | clear() on each iteration. This avoids allocating and deallocating all of DenseMap's memory on each iteration. llvm-svn: 52642
* Instead of adding an isSS field to LiveInterval to denote stack slot. Use ↵Evan Cheng2008-06-232-5/+4
| | | | | | top bit of 'reg' instead. If the top bit is set, than the LiveInterval represents a stack slot live interval. llvm-svn: 52639
* Duncan pointed out this code could be tidied.Dan Gohman2008-06-231-6/+3
| | | | llvm-svn: 52624
* Port some integer multiplication fixes from LegalizeDAG.Duncan Sands2008-06-231-8/+24
| | | | | | | Bail out with an error if there is no libcall available for the given size of integer. llvm-svn: 52622
* Support for expanding the result of EXTRACT_ELEMENT.Duncan Sands2008-06-234-7/+24
| | | | llvm-svn: 52621
* Cleanup up LegalizeTypes handling of loads andDuncan Sands2008-06-235-24/+22
| | | | | | stores. llvm-svn: 52620
* Make custom lowering of ADD work correctly. ThisDuncan Sands2008-06-221-3/+3
| | | | | | | | | | fixes PR2476; patch by Richard Osborne. The same problem exists for a bunch of other operators, but I'm ignoring this because they will be automagically fixed when the new LegalizeTypes infrastructure lands, since it already solves this problem centrally. llvm-svn: 52610
* Simplify some getNode calls.Dan Gohman2008-06-211-6/+3
| | | | llvm-svn: 52604
* canClobberPhysRegDefs shouldn't called without checking hasPhysRegDefs;Dan Gohman2008-06-211-3/+2
| | | | | | check this with an assert. llvm-svn: 52603
* Use clear() to zero an existing APInt.Dan Gohman2008-06-211-1/+1
| | | | llvm-svn: 52601
* Use back() instead of [size()-1].Dan Gohman2008-06-211-1/+1
| | | | llvm-svn: 52600
* Remove a redundant return.Dan Gohman2008-06-211-2/+0
| | | | llvm-svn: 52585
* Remove ScheduleDAG's SUnitMap altogether. Instead, use SDNode's NodeIdDan Gohman2008-06-214-50/+38
| | | | | | | field, which is otherwise unused after instruction selection, as an index into the SUnit array. llvm-svn: 52583
* Add a priority queue class, which is a wrapper around std::priority_queueDan Gohman2008-06-212-11/+8
| | | | | | | and provides fairly efficient removal of arbitrary elements. Switch ScheduleDAGRRList from std::set to this new priority queue. llvm-svn: 52582
* Support for load/store of expanded float types. IDuncan Sands2008-06-213-1/+65
| | | | | | | don't know if a truncating store is possible here, but added support for it anyway. llvm-svn: 52577
* Change ScheduleDAG's SUnitMap from DenseMap<SDNode*, vector<SUnit*> >Dan Gohman2008-06-214-34/+46
| | | | | | | | to DenseMap<SDNode*, SUnit*>, and adjust the way cloned SUnit nodes are handled so that only the original node needs to be in the map. This speeds up llc on 447.dealII.llvm.bc by about 2%. llvm-svn: 52576
* Undo spill weight tweak. Need to investigate the performance regressions.Evan Cheng2008-06-212-10/+5
| | | | llvm-svn: 52572
* Simplify some template parameterization.Dan Gohman2008-06-211-19/+12
| | | | llvm-svn: 52571
* Enhanced heuristic to determine the *best* register to spill. Instead of ↵Evan Cheng2008-06-201-66/+176
| | | | | | | | picking the register with the lowest spill weight. Consider (up to) 2 additional registers with spill weights that are close to the lowest spill weight. The one with fewest defs and uses that conflicts with the current interval (weighted by loop depth) is the spill candidate. This is not always a win, but there are much more wins than loses and wins tend to be more noticeable. llvm-svn: 52554
* Share some code that is common between integer andDuncan Sands2008-06-206-412/+560
| | | | | | float expansion (and sometimes vector splitting too). llvm-svn: 52548
* Rename the operation of turning a float type into anDuncan Sands2008-06-205-77/+76
| | | | | | | | | integer of the same type. Before it was "promotion", but this is confusing because it is quite different to promotion of integers. Call it "softening" instead, inspired by "soft float". llvm-svn: 52546
* Clean up some uses of std::distance, now that we have allnodes_size.Dan Gohman2008-06-202-4/+3
| | | | llvm-svn: 52545
* Teach ReturnInst lowering about aggregate return values.Dan Gohman2008-06-201-22/+28
| | | | llvm-svn: 52522
* Fix the index calculations for the extractvalue lowering code.Dan Gohman2008-06-201-2/+2
| | | | llvm-svn: 52517
* Simplify the ComputeLinearIndex logic and fix a few bugs.Dan Gohman2008-06-201-16/+10
| | | | llvm-svn: 52516
* ISD::UNDEF should be expanded recursively / iteratively.Evan Cheng2008-06-191-1/+0
| | | | llvm-svn: 52508
* Use the transferSuccessors helper function.Dan Gohman2008-06-191-5/+1
| | | | llvm-svn: 52495
* Missed a check.Evan Cheng2008-06-191-1/+1
| | | | llvm-svn: 52487
* Revert my last patch, which was causing regression test failures.Owen Anderson2008-06-191-78/+69
| | | | llvm-svn: 52485
* Coalesce copy from one register class to a sub register class. e.g. ↵Evan Cheng2008-06-192-25/+98
| | | | | | X86::MOV16to16_. llvm-svn: 52480
* Cosmetic changes.Evan Cheng2008-06-191-3/+5
| | | | llvm-svn: 52479
* Minor spiller tweak to unfavor reload into load/store instructions.Evan Cheng2008-06-191-4/+8
| | | | llvm-svn: 52477
* Insert empty slots into the instruction numbering in live intervals, so that ↵Owen Anderson2008-06-191-69/+78
| | | | | | | | we can more easily add new instructions. llvm-svn: 52475
* Fix the source line debug information for the Windows platform.Argyrios Kyrtzidis2008-06-181-5/+88
| | | | | | | According to DWARF-2 specification, the line information is provided through an offset in the .debug_line section. Replace the label reference that is used with a section offset. llvm-svn: 52468
* Complete support for two-address pass rematerialization. Now *almost* always ↵Evan Cheng2008-06-181-49/+128
| | | | | | a win. llvm-svn: 52452
* Cosmetic.Evan Cheng2008-06-181-1/+1
| | | | llvm-svn: 52450
* Live-through live interval is [mbb start, mbb end+1].Evan Cheng2008-06-171-1/+1
| | | | llvm-svn: 52431
* When extending a liveinterval by commuting, don't throw away the live ranges ↵Evan Cheng2008-06-171-2/+13
| | | | | | that are not affected. llvm-svn: 52430
* It's not safe to remove SUBREG_TO_REG that looks like identity copies, e.g. ↵Evan Cheng2008-06-171-11/+6
| | | | | | movl %eax, %eax on x86-64 actually does a zero-extend. llvm-svn: 52421
* Split type expansion into ExpandInteger and ExpandFloatDuncan Sands2008-06-177-1500/+1611
| | | | | | | | rather than bundling them together. Rename FloatToInt to PromoteFloat (better, if not perfect). Reorganize files by types rather than by operations. llvm-svn: 52408
OpenPOWER on IntegriCloud