- 8c20ec5 Enable support for vector sext and trunc: by Nadav Rotem · 14 years ago
- 8ca814c Merge information about the number of zero, one, and sign bits of live-out by Cameron Zwarich · 14 years ago
- 9b6af8d Add a getNumSignBits() method to APInt. by Cameron Zwarich · 14 years ago
- 324a24f Add a mechanism for invalidating the LiveOutInfo of a PHI, and use it whenever by Cameron Zwarich · 14 years ago
- a46cd97 Track blocks visited in reverse postorder. by Cameron Zwarich · 14 years ago
- e1497b9 Refactor the LiveOutInfo interface into a few methods on FunctionLoweringInfo by Cameron Zwarich · 14 years ago
- 2dbe285 Have isel visit blocks in reverse postorder rather than an undefined order. This by Cameron Zwarich · 14 years ago
- 4ecc82e Roll out r126169 and r126170 in an attempt to fix the selfhost bot. by Cameron Zwarich · 14 years ago
- 92efda7 Merge information about the number of zero, one, and sign bits of live-out registers by Cameron Zwarich · 14 years ago
- 63a8dae Have isel visit blocks in reverse postorder rather than an undefined order. This by Cameron Zwarich · 14 years ago
- 68e6bee Revert r124611 - "Keep track of incoming argument's location while emitting LiveIns." by Devang Patel · 14 years ago
- ed9b934 Fix 9267; Add vector zext support. by Nadav Rotem · 14 years ago
- a778f5c Do not lose debug info of an inlined function argument even if the argument is only used through GEPs. by Devang Patel · 14 years ago
- 1646953 Roll out r125794 to help diagnose the llvm-gcc-i386-linux-selfhost failure. by Cameron Zwarich · 14 years ago
- a4defc0 Do not lose debug info of an inlined function argument even if the argument is only used through GEPs. by Devang Patel · 14 years ago
- fbd53f7 Fix wrong logic in promotion of signed mul-with-overflow (I pointed this out at by Duncan Sands · 14 years ago
- a901129 Swap VT and DebugLoc operands of getExtLoad() for consistency with by Stuart Hastings · 14 years ago
- 7bccf6a Refactor zero folding slightly. Clean up todo. by Eric Christopher · 14 years ago
- 169e155 The change for PR9190 wasn't quite right. We need to avoid making the by Eric Christopher · 14 years ago
- 2ca5c86 convert ConstantVector::get to use ArrayRef. by Chris Lattner · 14 years ago
- 7583190 revert my ConstantVector patch, it seems to have made the llvm-gcc by Chris Lattner · 14 years ago
- 283c8ca Switch ConstantVector::get to use ArrayRef instead of a pointer+size by Chris Lattner · 14 years ago
- 593051b fix PR9210 by implementing some type legalization logic for by Chris Lattner · 14 years ago
- 463b3c2 fix two comment thinkos by Chris Lattner · 14 years ago
- 0a9481f Enhance ComputeMaskedBits to know that aligned frameindexes by Chris Lattner · 15 years ago
- e075118 Revisit my fix for PR9028: the issue is that DAGCombine was by Chris Lattner · 15 years ago
- 915eeb4 when legalizing extremely wide shifts, make sure that by Chris Lattner · 15 years ago
- d302773 fix visitShift to properly zero extend the shift amount if the provided operand by Chris Lattner · 15 years ago
- 609d54e A fix for 9165. by Nadav Rotem · 15 years ago
- f94fdb6 SimplifySelectOps can only handle selects with a scalar condition. Add a check by Nadav Rotem · 15 years ago
- 5a4552c by Nadav Rotem · 15 years ago
- 480d1e3 Remove comment about an argument that was removed couple of years ago. by Devang Patel · 15 years ago
- 92e9466 Introducing a new method of tracking register pressure. We can't by Andrew Trick · 15 years ago
- cd5af07 whitespace by Andrew Trick · 15 years ago
- 31959b1 Given a pair of floating point load and store, if there are no other uses of by Evan Cheng · 15 years ago
- 9a14a36 Take Bill Wendling's suggestion for structuring a couple of asserts. by Matt Beaumont-Gay · 15 years ago
- e9a7ea6 Keep track of incoming argument's location while emitting LiveIns. by Devang Patel · 15 years ago
- 4e3740e Fix bug where ReduceLoadWidth was creating illegal ZEXTLOAD instructions. by Richard Osborne · 15 years ago
- 9b108a3 Teach DAGCombine to fold fold (sra (trunc (sr x, c1)), c2) -> (trunc (sra x, c1+c2) when c1 equals the amount of bits that are truncated off. by Benjamin Kramer · 15 years ago
- 2c94b42 Add the missing sub identity "A-(A-B) -> B" to DAGCombine. by Benjamin Kramer · 15 years ago
- ed4efd3 Fix build with stdcxx by using llvm::next. Patch by Joerg Sonnenberger! by Nick Lewycky · 15 years ago
- e787ddb Remove a temporary workaround for a lencod miscompile. Depends on the fix in r124442. by Andrew Trick · 15 years ago
- 7e7fc10 Speculatively revert r124380. by Devang Patel · 15 years ago
- bfae5f3 While legalizing SDValues do not drop SDDbgValues, trasfer them to new legal nodes. by Devang Patel · 15 years ago
- 4789aca Try harder to not have unused variables. by Matt Beaumont-Gay · 15 years ago
- a8af137 Opt-mode -Wunused-variable cleanup by Matt Beaumont-Gay · 15 years ago
- 39078a8 Reapply 124301 by Devang Patel · 15 years ago
- 6b9a293 Initialize variable to get rid of clang warning. by Bill Wendling · 15 years ago
- 4a95c19 Revert 124301. by Devang Patel · 15 years ago
- c47fd9f Revert r124302 by Devang Patel · 15 years ago
- cfe33c4 by David Greene · 15 years ago
- d2c9793 While legalizing SDValues do not drop SDDbgValues, trasfer them to new legal nodes. by Devang Patel · 15 years ago
- 6f121fd Process valid SDDbgValues even if the node does not have any order assigned. by Devang Patel · 15 years ago
- 55d20e8 Refactor. by Devang Patel · 15 years ago
- 9158509 by David Greene · 15 years ago
- a2e868d Provide an interface to transfer SDDbgValue from one SDNode to another. by Devang Patel · 15 years ago
- 8f31428 by Devang Patel · 15 years ago
- 3a00ffa This assertion is too restrictive, it does not apply for dangling dbg value nodes (nodes where dbg.value intrinsic preceds use of the value). by Devang Patel · 15 years ago
- a3ee3ef Speculatively revert r124138. by Devang Patel · 15 years ago
- 224a180 Resolve DanglingDbgValue of PHI nodes where the use follows dbg.value intrinisic. by Devang Patel · 15 years ago
- c48d50f Temporarily workaround JM/lencod miscompile (SIGSEGV). rdar://problem/8893967 by Andrew Trick · 15 years ago
- 584520e Null initialize a few variables flagged by by Ted Kremenek · 15 years ago
- d1dace8 Enable support for precise scheduling of the instruction selection by Andrew Trick · 15 years ago
- c8bfd1d Convert -enable-sched-cycles and -enable-sched-hazard to -disable by Andrew Trick · 15 years ago
- 6214373 My editor's indent went crazy. Fix. by Eric Christopher · 15 years ago
- 0205098 Expand invalid return values for umulo and smulo. Handle these similarly by Eric Christopher · 15 years ago
- 0bc3086 Selection DAG scheduler register pressure heuristic fixes. by Andrew Trick · 15 years ago
- e338581 Use only one API at a time. by Eric Christopher · 15 years ago
- 38a1826 If we can, lower the multiply part of a umulo/smulo call to a libcall by Eric Christopher · 15 years ago
- 955ed73 Remove unused variables found by gcc-4.6's -Wunused-but-set-variable. by Jeffrey Yasskin · 15 years ago
- f224322 Remove checking that prevented overlapping CALLSEQ_START/CALLSEQ_END by Stuart Hastings · 15 years ago
- c9b6a3e Fix an off-by-one error in ctpop combining. by Benjamin Kramer · 15 years ago
- d822892 Add a DAGCombine to turn (ctpop x) u< 2 into (x & x-1) == 0. by Benjamin Kramer · 15 years ago
- b99fdee reapply my fix for PR8961 with a tweak to properly handle by Chris Lattner · 15 years ago
- 5df5a22 Add an assert so we don't silently miscompile ctpop for bit widths > 128. by Benjamin Kramer · 15 years ago
- b6516ae Reimplement CTPOP legalization with the "best" algorithm from by Benjamin Kramer · 15 years ago
- ca5f616 Delete an assignment to ThisBB which isn't needed, and tidy up some comments. by Dan Gohman · 15 years ago
- f697c8a Support for precise scheduling of the instruction selection DAG, by Andrew Trick · 15 years ago
- 67d9891 Set the insertion point correctly for instructions generated by load folding: by Chris Lattner · 15 years ago
- dd11ea4 Fix r123346 to handle scalar types too. by Dan Gohman · 15 years ago
- b866543 Apply the patch from PR8958, which allows llc to get slightly by Dan Gohman · 15 years ago
- 93c7042 Move ExpandAtomic into the integer expansion routines - it's only used there. by Eric Christopher · 15 years ago
- 97fd9a5 Fix PR 8916 (qv for analysis), at least the immediate problem. by Dale Johannesen · 15 years ago
- 16c29b5 Rename TargetFrameInfo into TargetFrameLowering. Also, put couple of FIXMEs and fixes here and there. by Anton Korobeynikov · 15 years ago
- c9df025 Simplify a bunch of isVirtualRegister() and isPhysicalRegister() logic. by Jakob Stoklund Olesen · 15 years ago
- 4314268 Replace TargetRegisterInfo::printReg with a PrintReg class that also works without a TRI instance. by Jakob Stoklund Olesen · 15 years ago
- 358de24 Use an IndexedMap for LiveOutRegInfo to hide its dependence on TargetRegisterInfo::FirstVirtualRegister. by Jakob Stoklund Olesen · 15 years ago
- c36b706 Do not model all INLINEASM instructions as having unmodelled side effects. by Evan Cheng · 15 years ago
- 5e8b833 Add ARM patterns to match EXTRACT_SUBVECTOR nodes. by Bob Wilson · 15 years ago
- 6736e19 Change EXTRACT_SUBVECTOR to require a constant index. by Bob Wilson · 15 years ago
- d9aa800 Fix the other problem reported in PR8582. Testcase and patch by Nadav Rotem. by Duncan Sands · 15 years ago
- 8d93d19 Add some fairly duplicated code to let type legalization split illegal by Eric Christopher · 15 years ago
- 0521928 Re-implement r122936 with proper target hooks. Now getMaxStoresPerMemcpy by Evan Cheng · 15 years ago
- 255874f Revert r122936. I'll re-implement the change. by Evan Cheng · 15 years ago
- 9a9d847 r105228 reduced the memcpy / memset inline limit to 4 with -Os to avoid blowing by Evan Cheng · 15 years ago
- d08e5b4 Avoid zero extend bit test operands to pointer type if all the masks fit in by Evan Cheng · 15 years ago
- 0b71d39 Optimize: by Evan Cheng · 15 years ago
- 762a17a 80-cols. by Eric Christopher · 15 years ago
- 2d31d14 Remove TODO, these appear to be implemented. by Eric Christopher · 15 years ago
- 8022036 Try to reuse the value when lowering memset. by Benjamin Kramer · 15 years ago