- fadc999 The trampoline intrinsic can be marked IntrWriteArgMem, by Duncan Sands · 16 years ago
- 054a40c IntrWriteArgMem implies nocapture, but this wasn't by Duncan Sands · 16 years ago
- b5aa3ad The x86-64 red zone is now being used. by Dan Gohman · 16 years ago
- aa03649 Extend the IndVarSimplify support for promoting induction variables: by Dan Gohman · 16 years ago
- f108e2e Clarify debug output. by Dan Gohman · 16 years ago
- cd3eb9b Simplify some code. hasComputableLoopEvolution is overkill in this case. by Dan Gohman · 16 years ago
- b463272 Testcase for llvm-gcc 64510. by Dale Johannesen · 16 years ago
- c7ce29b Teach x86 target -soft-float. by Evan Cheng · 16 years ago
- 03ce042 In CodeGenPrepare's debug output, use WriteAsOperand instead of by Dan Gohman · 16 years ago
- e698696 Complete the sentance in this comment. I have reservations by Dan Gohman · 16 years ago
- 070fbe6 Correct a comment: since AccessArguments is now by Duncan Sands · 16 years ago
- ddd8c45 If a function only reads memory, then we know that... by Duncan Sands · 16 years ago
- 02b6a6a Mark strto* as readonly when the endptr is null. by Nick Lewycky · 16 years ago
- 4c49841 On strtod and friends, mark 'endptr' nocapture in the function prototype, and by Nick Lewycky · 16 years ago
- e70a683 Reapply r64301. These uses of "inline" can cause strange by Duncan Sands · 16 years ago
- e2b3fdd Switch from new[] + delete[] to malloc + free since llvm does not catch C++ exceptions. by Evan Cheng · 16 years ago
- 726135a If new[] fails, return 0 rather then trying to dereference a null pointer. by Evan Cheng · 16 years ago
- 8968a07 Reapply r64300: by Nick Lewycky · 16 years ago
- 826f7ce BasicAA was making the assumption that a local allocation which hadn't escaped by Nick Lewycky · 16 years ago
- 9194e8b Fix the code that checked if a SCEVAddRecExpr Start contains an by Dan Gohman · 16 years ago
- a200c8f Remove non-DebugLoc versions of BuildMI. "I got blisters on my fingers." by Dale Johannesen · 16 years ago
- 9481757 Remove non-DebugLoc versions of BuildMI from IA64, Mips. by Dale Johannesen · 16 years ago
- 8d13f8f Remove non-DebugLoc versions of BuildMI from X86. by Dale Johannesen · 16 years ago
- 0b4581f missed file by Dale Johannesen · 16 years ago
- d552eee Remove non-DebugLoc versions of buildMI from Sparc. by Dale Johannesen · 16 years ago
- 01b36e6 Remove non-DebugLoc versions of BuildMI from Alpha and Cell. by Dale Johannesen · 16 years ago
- bd9ef18 Remove refs to non-DebugLoc version of BuildMI from XCore, PIC16. by Dale Johannesen · 16 years ago
- 536a2f1 Remove refs to non-DebugLoc version of BuildMI from PowerPC. by Dale Johannesen · 16 years ago
- 8a43d98 and one more file by Dale Johannesen · 16 years ago
- b672840 Remove refs to non-DebugLoc versions of BuildMI from ARM. by Dale Johannesen · 16 years ago
- 92c1e12 Revert this. It was breaking stuff. by Bill Wendling · 16 years ago
- 54fc7d6 Turn off the old way of handling debug information in the code generator. Use by Bill Wendling · 16 years ago
- bc51172 Fix LSR's IV sorting function to explicitly sort by bitwidth by Dan Gohman · 16 years ago
- b445d74 Avoid order files for now, as they aren't supported in 3.79. by Mike Stump · 16 years ago
- c5b5051 Check in missing file. by Dale Johannesen · 16 years ago
- 21b5541 Eliminate a couple of non-DebugLoc BuildMI variants. Modify callers. by Dale Johannesen · 16 years ago
- c2390b1 Teach IndVarSimplify to optimize code using the C "int" type for by Dan Gohman · 16 years ago
- 0f123cf Add suppport for ConstantExprs of shufflevectors whose result type is not equal to the by Nate Begeman · 16 years ago
- 78e3e52 Arrange to print constants that match "n" and "i" constraints by Dale Johannesen · 16 years ago
- a971b29 Rename bugpoint's error message file so that if it somehow by Dan Gohman · 16 years ago
- c833246 Add a utility function to LoopInfo to return the exit block by Dan Gohman · 16 years ago
- 8c4d1b2 fix PR3538 for ARM. by Chris Lattner · 16 years ago
- f902d24 fix PR3538 for PPC by Chris Lattner · 16 years ago
- 1823c92 fix the X86 backend to just drop llvm.declare nodes for VLAs instead of by Chris Lattner · 16 years ago
- 4ed76e7 Adjust the sizes for a few SmallVectors to reflect their usage. by Dan Gohman · 16 years ago
- 14c4c1e make fast isel fall back to selectiondags for VLA llvm.declare intrinsics. by Chris Lattner · 16 years ago
- 5b22efa Add operator->, patch by Ben Laurie! by Chris Lattner · 16 years ago
- 327e98f add PR by Chris Lattner · 16 years ago
- d0ffa73 This code doesn't actually use the ExitingBlocks list. by Dan Gohman · 16 years ago
- 8679119 It's (currently) not safe to keep certain physical registers live across basic blocks, e.g. x86 fp stack registers. by Evan Cheng · 16 years ago
- f2b1471 Oops. Last second clean up messed things up. by Evan Cheng · 16 years ago
- 6d209c4 If availability info is kept when fallthrough into a bb, add the available registers to live-in set. by Evan Cheng · 16 years ago
- 8182347 Replace one of burr scheduling heuristic with something more sensible. Now calcMaxScratches simply compute the number of true data dependencies. This actually improve a couple of tests in dejagnu suite as many tests in llvm nightly test suite. by Evan Cheng · 16 years ago
- e0a056b Fix MingW build: define GTEST_OS_WINDOWS if OS is MingW, but disable exceptions. by Julien Lerouge · 16 years ago
- 9430fe7 Fix MingW build, patch by Kenneth Boyd! by Julien Lerouge · 16 years ago
- af9985c Fix a nasty bug (PR3550) where the inline pass could incorrectly mark by Chris Lattner · 16 years ago
- 0aefc0e improve naming of values in GVN, patch by Jay Foad! by Chris Lattner · 16 years ago
- 65c02fb fix PR3537: if resetting bbi back to the start of a block, we need to by Chris Lattner · 16 years ago
- afa7743 rename test to avoid messing with tab completion of dates. by Chris Lattner · 16 years ago
- da5eb71 Add method 'isSingleton()' to ImmutableSet. This returns true if the set contains exactly one element. by Ted Kremenek · 16 years ago
- b7c9eee Don't mark all args to strtod and friends as nocapture. by Nick Lewycky · 16 years ago
- d1c321a Move debug loc info along when the spiller creates new instructions. by Bill Wendling · 16 years ago
- f412f7c Apparently some MachineBasicBlock's don't have corresponding llvm basic blocks. by Evan Cheng · 16 years ago
- ce7f5a1 Remove a bogus assertion. It's possible a live-in available value is used by a previous instruction. by Evan Cheng · 16 years ago
- 7b25467 the two non-mask arguments to a shufflevector must be the same width, but they do not have to be the same by Nate Begeman · 16 years ago
- bd5e5a8 Edit description of floating point constants to by Dale Johannesen · 16 years ago
- 6cd093b Fill in a glaring omission in derived User classes, namely by Gabor Greif · 16 years ago
- fbef310 Propagate DebugLoc info for spiller call-backs. by Bill Wendling · 16 years ago
- c7fa16f Fix a comment. by Dan Gohman · 16 years ago
- c8db34c Don't use special heuristics for nodes with no data predecessors by Dan Gohman · 16 years ago
- 8bb332d Add FoldingSet profile method for bools. Patch by Ben Laurie! by Ted Kremenek · 16 years ago
- dce5e74 Update to match space changes in .ll by Daniel Dunbar · 16 years ago
- cc994ea Delete the heuristic for non-livein CopyFromReg nodes. Non-liveinness by Dan Gohman · 16 years ago
- d293e0d Don't try to set an EFLAGS operand to dead if no instruction was created. by Dan Gohman · 16 years ago
- 85b0ede Make a transformation added in 63266 a bit less aggressive. by Dale Johannesen · 16 years ago
- 905c7e9 Revert r64300 and r64301. These were causing the following errors respectively: by Bill Wendling · 16 years ago
- 535e3e8 Revert r64299: it breaks the build when configured by Duncan Sands · 16 years ago
- b91a2a1 These uses of "inline" can cause strange link-time by Duncan Sands · 16 years ago
- af06875 Make sure the SCC pass manager initializes any contained by Duncan Sands · 16 years ago
- 46dec95 Try this. Darwin -> LTO, PIC -> LTO + possibly gold too. by Nick Lewycky · 16 years ago
- 752272a Implement PR3495: local spiller optimization. The local spiller can now keep availability information over BB boundaries. It visits BB's in depth first order. After visiting a BB if it find a successor which has a single predecessor it visits the successor next without clearing the availability information. This allows the successor to omit reloads or change them into copies. by Evan Cheng · 16 years ago
- 47ac0f0 When scheduling a block in parts, keep track of the overall by Dan Gohman · 16 years ago
- ca70533 62987 disables LTO build on darwin. by Devang Patel · 16 years ago
- bd75b83 If llvm.dbg.region.end is disappearing then remove corresponding llvm.dbg.func.start also. by Devang Patel · 16 years ago
- 7dfa07f Don't print extra spaces in vector and array constants. This makes by Dan Gohman · 16 years ago
- 18e9a6b It isn't necessary to explicitly clear the contents of std::vector by Dan Gohman · 16 years ago
- 6919757 Don't try to strip a file we just moved. by Bill Wendling · 16 years ago
- 5cffa6f Use iterators to iterate through the Preds array instead of by Dan Gohman · 16 years ago
- bed353d Consider any instruction that modifies the stack pointer to be by Dan Gohman · 16 years ago
- 9e64bbb Factor out more code for computing register live-range informationfor by Dan Gohman · 16 years ago
- 0464a14 Ignore dbg intrinsic while folding unconditional branch. by Devang Patel · 16 years ago
- 1d76864 Handle llvm.x86.sse2.maskmov.dqu in 64-bit. by Evan Cheng · 16 years ago
- 7403eea 80 col violations. by Evan Cheng · 16 years ago
- f5f0a1f This is now done using a real i33, rather than by Duncan Sands · 16 years ago
- 7afe8fa Use early exits. Reduce indentation. by Devang Patel · 16 years ago
- d4e1845 Use const, to support platforms where strrchr returns a const char *. by Dan Gohman · 16 years ago
- cfbcd59 Answer a common FAQ: by Dan Gohman · 16 years ago
- 35b5107 Pacify gcc-4.3 (thinks IsConstant may be used uninitialized). by Duncan Sands · 16 years ago
- f0e366a Fix PR3457: Ignore control successors when looking for closest scheduled successor. A control successor doesn't read result(s) produced by the scheduling unit being evaluated. by Evan Cheng · 16 years ago
- f66d7b5 Do not clone llvm.dbg.func.start and corresponding llvm.dbg.region.end during inlining. by Devang Patel · 16 years ago