- c6feeb7 These should not have been committed. by Evan Cheng · 15 years ago
- 48eb39b no need to run llvm-as here. by Chris Lattner · 15 years ago
- 6e5dfd4 Instcombine constant folding can normalize gep with negative index to index with large offset. When instcombine objsize checking transformation sees these geps where the offset seemingly point out of bound, it should just return "i don't know" rather than asserting. by Evan Cheng · 15 years ago
- 0e488b3 Canonicalize ConstantInts to the right operand of commutative operators. by Dan Gohman · 15 years ago
- fc90d4c Actually enable the -enable-unsafe-fp-math tests. by Dan Gohman · 15 years ago
- 9265275 Mark the return address stack slot as mutable when moving the return address by Arnold Schwaighofer · 15 years ago
- 8ce05da Remove the logic for reasoning about NaNs from the code that forms by Dan Gohman · 15 years ago
- 278f958 When emitting an instruction which depends on both a post-incremented by Dan Gohman · 15 years ago
- 728ea19 add some no-unwinds, other minor cleanups. by Chris Lattner · 15 years ago
- 25e0f91 add a triple so that this doesn't fail due to linux/ppc register printing by Chris Lattner · 15 years ago
- ad14fcc filecheckize and add nouwinds. by Chris Lattner · 15 years ago
- 455080f IT turns out that during jumpless setcc lowering eq and ne were swapped. by Anton Korobeynikov · 15 years ago
- f1c6428 fix and un-xfail X86/vec_ss_load_fold.ll by Chris Lattner · 15 years ago
- 7f6de7d temporarily disable this. by Chris Lattner · 15 years ago
- aae01f1 Check for overflow when scaling up an add or an addrec for scaled reuse. by Dan Gohman · 15 years ago
- 5dfa267 Add support for the 'alignstack' attribute to the x86 backend. Fixes PR5254. by Charles Davis · 15 years ago
- 3eba667 Revert commits 96556 and 96640, because commit 96556 breaks the by Duncan Sands · 15 years ago
- 97a35fc Transform (xor (setcc), (setcc)) == / != 1 to (xor (setcc), (setcc)) != / == 1. by Evan Cheng · 15 years ago
- 1b7bf18 When determining the set of interesting reuse factors, consider by Dan Gohman · 15 years ago
- 1c40e18 getSplatIndex assumes that the first element of the mask contains the splat index by Mon P Wang · 15 years ago
- 352d352 Always normalize spill weights, also for intervals created by spilling. by Jakob Stoklund Olesen · 15 years ago
- 49d7f8d Make CodePlacementOpt detect special EH control flow by by Dan Gohman · 15 years ago
- 7954926 remove empty file by Chris Lattner · 15 years ago
- 9f6c4c1 Use NEON vmin/vmax instructions for floating-point selects. Radar 7461718. by Bob Wilson · 15 years ago
- 89bb7b5 Some dag combiner goodness: by Evan Cheng · 15 years ago
- 2a5431e Don't check for comments, which vary between subtargets. by Dan Gohman · 15 years ago
- 968cb93 Don't attempt to divide INT_MIN by -1; consider such cases to have overflowed. by Dan Gohman · 15 years ago
- ca0aacc roundss is an sse 4 thing, fix the test on non-sse41 builders by Chris Lattner · 15 years ago
- 4a12de7 Make g5 target explicit; scheduling affects register choice. by Dale Johannesen · 15 years ago
- 92d3ada fix rdar://7653908, a crash on a case where we would fold a load by Chris Lattner · 15 years ago
- 55f9adf Adjust register numbers in tests to compensate for the new lack of R2. by Dale Johannesen · 15 years ago
- 5583b83 filecheckize by Chris Lattner · 15 years ago
- ae3ecf9 Look for SSE and instructions of this form: (and x, (build_vector c1,c2,c3,c4)). by Evan Cheng · 15 years ago
- 8939b0d by David Greene · 15 years ago
- 1665b0a Fix pr6111: Avoid using the LR register for the target address of an indirect by Bob Wilson · 15 years ago
- c2385a0 Split the main for-each-use loop again, this time for GenerateTruncates, by Dan Gohman · 15 years ago
- 9184b25 Preliminary patch to improve dwarf EH generation - Hooks to return Personality / FDE / LSDA / TType encoding depending on target / options (e.g. code model / relocation model) - MCIzation of Dwarf EH printer to use encoding information - Stub generation for ELF target (needed for indirect references) - Some other small changes here and there by Anton Korobeynikov · 15 years ago
- 8eea48a Fix PR6300. by Jakob Stoklund Olesen · 15 years ago
- f1b0a34 Last week we were generating code with duplicate induction variables in this by Bob Wilson · 15 years ago
- ebd024f remove empty file. by Chris Lattner · 15 years ago
- cd10d14 revert r96241. It breaks two regression tests, isn't documented, by Chris Lattner · 15 years ago
- 37914c8 fix PR6305 by handling BlockAddress in a helper function by Chris Lattner · 15 years ago
- 786da7d by David Greene · 15 years ago
- d135f14 Fix PR6283. by Jakob Stoklund Olesen · 15 years ago
- bf9b221 Besides removing phi cycles that reduce to a single value, also remove dead by Bob Wilson · 15 years ago
- c12da8d When save/restoring CR at prolog/epilog, in a large by Dale Johannesen · 15 years ago
- c8f6326 Testcases for recent stdcall / fastcall mangling improvements by Anton Korobeynikov · 15 years ago
- 4dd162f Cleanup stdcall / fastcall name mangling. by Anton Korobeynikov · 15 years ago
- 572645c Reapply the new LoopStrengthReduction code, with compile time and by Dan Gohman · 15 years ago
- fe61fb1 Add a new pass on machine instructions to optimize away PHI cycles that by Bob Wilson · 15 years ago
- 4a540f0 Reapply coalescer fix for better cross-class coalescing. by Jakob Stoklund Olesen · 15 years ago
- f7ea6c3 The previous fix of widening divides that trap was too fragile as it depends on custom by Mon P Wang · 15 years ago
- 5e2b05a Delete dead PHI machine instructions. These can be created due to type by Bob Wilson · 15 years ago
- 046632f Now that ShrinkDemandedOps() is separated out from DAG combine. It sometimes leave some obvious nops which dag combine used to clean up afterwards e.g. (trunk (ext n)) -> n. Look for them and squash them. by Evan Cheng · 15 years ago
- cb921e7 move tests that depend on the x86 backend out of codegen/generic, by Chris Lattner · 15 years ago
- 4a26cc4 make target independent. by Chris Lattner · 15 years ago
- b00cd1c merge a target-specific add test into x86 directory. by Chris Lattner · 15 years ago
- 7816239 merge another test in, drop the trivially constant folded cases. by Chris Lattner · 15 years ago
- 71d845c consolidate and filecheckize two tests. by Chris Lattner · 15 years ago
- 2b7ff1b merge two tests, make target independent. by Chris Lattner · 15 years ago
- d863b76 convert to filecheck. by Chris Lattner · 15 years ago
- 25ceb5f add an x86 implementation of MCTargetExpr for by Chris Lattner · 15 years ago
- 618f177 When CodeGen'ing unoptimized code, there may be unfolded constant expressions by Dan Gohman · 15 years ago
- c2e93b2 In guaranteed tailcall mode, don't decline the tailcall optimization by Dan Gohman · 15 years ago
- 00a99a3 Run codegen dce pass for all targets at all optimization levels. Previously it's by Evan Cheng · 15 years ago
- e3e86dc Remove a large test case that (soon will) no longer make sense. by Evan Cheng · 15 years ago
- 2f82ca9 Fix alignment on ppc linux. This fixes the build of crtend.o by Rafael Espindola · 15 years ago
- f22f9b3 Do not emit callseq instructions around sibcalls. This eliminated some unnecessary stack adjustments. by Evan Cheng · 15 years ago
- e6373eb Handle AddrMode6 (for NEON load/stores) in Thumb2's rewriteT2FrameIndex. by Bob Wilson · 15 years ago
- d450e5b Don't unroll loops containing function calls. by Jakob Stoklund Olesen · 15 years ago
- 63d58eb Make test more fucused eliminating extraneous bits. by Bill Wendling · 15 years ago
- 401f030 Fix test. by Evan Cheng · 15 years ago
- 5f94193 Handle tail call with byval arguments. by Evan Cheng · 15 years ago
- 5ba8bf6 When the scheduler unfold a load folding instruction it move some of the predecessors to the unfolded load. It decides what gets moved to the load by checking whether the new load is using the predecessor as an operand. The check neglects the cases whether the predecessor is a flagged scheduling unit. by Evan Cheng · 15 years ago
- 744f579 An empty global constant (one of size 0) may have a section immediately by Bill Wendling · 15 years ago
- e4d2d96 Fix small bug in handling instructions with more than one implicitly defined operand. by Jakob Stoklund Olesen · 15 years ago
- 1f2fda9 Re-enable x86 tail call optimization. by Evan Cheng · 15 years ago
- 87c06d6 add support for the sparcv9-*-* target triple to turn on by Chris Lattner · 15 years ago
- 8148ae8 Speculatively disable x86 automatic tail call optimization while we track down a self-hosting issue. by Evan Cheng · 15 years ago
- 394f1b5 Make test less fragile by Evan Cheng · 15 years ago
- febc816 Revert 94937 and move the noreturn check to codegen. by Evan Cheng · 15 years ago
- 86809cc Allow all types of callee's to be tail called. But avoid automatic tailcall if the callee is a result of bitcast to avoid losing necessary zext / sext etc. by Evan Cheng · 15 years ago
- 9a6636b Reapply 95050 with a tweak to check the register class. by Dale Johannesen · 15 years ago
- e0bb20c make these less sensitive to asm verbose changes by disabling it for them. by Chris Lattner · 15 years ago
- 62a0413 Test revert 95050; there's a good chance it's causing buildbot failure. by Dale Johannesen · 15 years ago
- b2c9290 Perform sibcall in some cases when arguments are passes memory. Look for cases by Evan Cheng · 15 years ago
- 5ce0ee9 Make local RA smarter about reusing input register of a copy by Dale Johannesen · 15 years ago
- 7276c8c Fix PR6196. GV callee may not be a function. by Evan Cheng · 15 years ago
- a00ef11 Update this test for a trivial register allocation difference. by Dan Gohman · 15 years ago
- 0a4fd46 Undo r94946 now all the tests are passing again. by Evan Cheng · 15 years ago
- 843bd69 Avoid recursive sibcall's. by Evan Cheng · 15 years ago
- 90cfc13 Fix a gross typo: ARMv6+ may or may not support unaligned memory operations. by Anton Korobeynikov · 15 years ago
- a6bff98 Allow more tailcall optimization: calls with inputs that are all passed in registers. by Evan Cheng · 15 years ago
- 7096ae4 Catch more trivial tail call opportunities: no inputs and output types match. by Evan Cheng · 15 years ago
- a34103f convert the last 3 targets to use EmitFunctionBody() now that by Chris Lattner · 15 years ago
- 5fef8bc Fix a bug introduced by r94490 where it created a X86ISD::CMP whose output type is different from its inputs. by Evan Cheng · 15 years ago
- d49fe1b Give AsmPrinter the most common expected implementation of by Chris Lattner · 15 years ago
- 10e7c60 emit a 0 byte instead of a noop if a function is empty on darwin. by Chris Lattner · 15 years ago
- 6f03433 Quick fix to a test that is currently failing on every Linux build bot. No idea by Chandler Carruth · 15 years ago
- b171245 Perform trivial tail call optimization for callees with "C" ABI. These are done by Evan Cheng · 15 years ago