- 56f2987 r70270 isn't ready yet. Back this out. Sorry for the noise. by Bill Wendling · 17 years ago
- d0ae159 Massive check in. This changes the "-fast" flag to "-O#" in llc. If you want to by Bill Wendling · 17 years ago
- 8d6d4b9 2nd attempt, fixing SSE4.1 issues and implementing feedback from duncan. by Nate Begeman · 17 years ago
- b93db66 Revert 69952. Causes testsuite failures on linux x86-64. by Rafael Espindola · 17 years ago
- bb881d6 PR2957 by Nate Begeman · 17 years ago
- 9478c3f Factor out a bit of code that appears in several places into a utility function. by Dan Gohman · 17 years ago
- a290ab4 Handle Void types in ComputeValueVTs. This doesn't currently occur, by Dan Gohman · 17 years ago
- ad968ee Inline asm's were still introducing bogus dependencies; by Dale Johannesen · 17 years ago
- dab01f3 Do not treat beginning of inlined scope as beginning of normal function scope if the location info is missing. by Devang Patel · 17 years ago
- 9ac4390 Record line number at the beginning of a func.start. by Devang Patel · 17 years ago
- 653dee0 In -fast mode do what FastISel does. This code could use some refactoring help! by Devang Patel · 17 years ago
- 83593f4 Do not force asm's to be chained if they don't touch by Dale Johannesen · 17 years ago
- 097f630 Make these errors more noticable in build logs. by Daniel Dunbar · 17 years ago
- 0431504 Right now, Debugging information to encode scopes (DW_TAG_lexical_block) relies on DBG_LABEL. Unfortunately this intefers with the quality of optimized code. by Devang Patel · 17 years ago
- 80be351 by Devang Patel · 17 years ago
- 825236b Revert r68847. It breaks the build on non-Darwin targets, with this message by Dan Gohman · 17 years ago
- 790e609 Keep track of inlined functions and their locations. This information is collected when nested llvm.dbg.func.start intrinsics are seen. (Right now, inliner removes nested llvm.dbg.func.start intrinisics during inlining.) by Devang Patel · 17 years ago
- de912e2 Remove the obsolete SelectionDAG::getNodeValueTypes and simplify by Dan Gohman · 17 years ago
- 0e8d199 Generalize ExtendUsesToFormExtLoad to be usable for ANY_EXTEND, by Dan Gohman · 17 years ago
- e6db8ca Don't copy the operand of a SwitchInst into virtual registers as by Dan Gohman · 17 years ago
- 10f7c3d Revert prev. patch for now. by Devang Patel · 17 years ago
- ddafc03 Right now DBG_LABEL are required for llvm.dbg.region_start and llvm.dbg.region_end in non-fast mode also. by Devang Patel · 17 years ago
- 97a20b8 Fix live-out reg logic to not insert over-aggressive AssertZExt by Dan Gohman · 17 years ago
- 2e9f42b Revert 67132. This is breaking some objective-c apps. by Evan Cheng · 17 years ago
- eb1646d When optimizing with debug info, don't keep the by Dale Johannesen · 17 years ago
- c35847e more tidying: name the components of PhysReg in the case when by Chris Lattner · 17 years ago
- 42eceb3 Tidy a bit more. by Chris Lattner · 17 years ago
- 246eda4 simplify this code a bit now that "allocation to a vreg class" can never fail. by Chris Lattner · 17 years ago
- 968c3b0 Model inline asm constraint which ties an input to an output register as machine operand TIED_TO constraint. This eliminated the need to pre-allocate registers for these. This also allows register allocator can eliminate the unneeded copies. by Evan Cheng · 17 years ago
- 2e55923 For inline asm output operand that matches an input. Encode the input operand index in the high bits. by Evan Cheng · 17 years ago
- 4606b12 Don't force promotion of return arguments on the callee. by Rafael Espindola · 17 years ago
- 2363d0b Fix codegen to compute the size of an allocation by multiplying the by Chris Lattner · 17 years ago
- fa54bc2 Oops...I committed too much. by Bill Wendling · 17 years ago
- b02eadf Temporarily XFAIL this test. by Bill Wendling · 17 years ago
- c6869f4 Pass in a std::string when getting the names of debugging things. This cuts down by Bill Wendling · 17 years ago
- a49de9d Revert BuildVectorSDNode related patches: 65426, 65427, and 65296. by Evan Cheng · 17 years ago
- 7d12ea0 Fix big-endian codegen bug. We're splitting up by Dale Johannesen · 17 years ago
- 86673f2 Clean up dwarf writer, part 1. This eliminated the horrible recursive getGlobalVariablesUsing and replaced it something readable. It eliminated use of slow UniqueVector and replaced it with StringMap, SmallVector, and DenseMap, etc. It also fixed some non-deterministic behavior. by Evan Cheng · 17 years ago
- c5437ea Overhaul my earlier submission due to feedback. It's a large patch, but most of by Bill Wendling · 17 years ago
- 786c597 - Use the "Fast" flag instead of "OptimizeForSize" to determine whether to emit by Bill Wendling · 17 years ago
- 9d31aca Introduce the BuildVectorSDNode class that encapsulates the ISD::BUILD_VECTOR by Scott Michel · 17 years ago
- 7b9f38a Temporarily revert r65065. It was causing test failures. by Bill Wendling · 17 years ago
- df78dcc Check for -fast here too. by Bill Wendling · 17 years ago
- 19e0a5b Generate these labels when we're in "fast" mode, not simply when we're no in by Bill Wendling · 17 years ago
- cf0da6c Remove trailing whitespace to reduce later commit patch noise. by Scott Michel · 17 years ago
- 3c50922 --- Merging (from foreign repository) r64714 into '.': by Bill Wendling · 17 years ago
- d1e179d Unbreak the build on win32. Cleanup some warning. by Cedric Venet · 17 years ago
- 65c0fd4 Revert this. It was breaking stuff. by Bill Wendling · 17 years ago
- 1c21ac3 Turn off the old way of handling debug information in the code generator. Use by Bill Wendling · 17 years ago
- fe174df Simplify code. by Evan Cheng · 17 years ago
- 020588c Make sure constant subscript is truncated to ptr size if it may not fit. by Evan Cheng · 17 years ago
- 8493575 Remove more non-DebugLoc getNode variants. Use by Dale Johannesen · 17 years ago
- 03c34d0 Clear out the CurDebugLoc info when doing a 'clear' on the SDL object. by Bill Wendling · 17 years ago
- b842d52 Reapply 63765. Patches for clang and llvm-gcc to follow. by Dale Johannesen · 17 years ago
- 7ae8c8b Remove non-DebugLoc versions of getMergeValues, ZeroExtendInReg. by Dale Johannesen · 17 years ago
- ae616c2 Reverting 63765. This broke the build of both clang and llvm-gcc. by Dale Johannesen · 17 years ago
- 6ae3aa8 New feature: add support for target intrinsics being defined in the by Nate Begeman · 17 years ago
- f2bb6f0 Remove some more non-DebugLoc versions of construction by Dale Johannesen · 17 years ago
- 8526388 Remove a few non-DebugLoc versions of node creation functions. by Dale Johannesen · 17 years ago
- 3a09f55 DebugLoc propagation; adjustment to things omitted from SelectionDagBuild. by Dale Johannesen · 17 years ago
- 135227a Pass in something sensible for the debug location information when creating the by Bill Wendling · 17 years ago
- fa50a23 Explicitly pass in the "unknown" debug location. This is probably not by Bill Wendling · 17 years ago
- e3c7836 Create DebugLoc information in FastISel. Several temporary methods were by Bill Wendling · 17 years ago
- 4182603 Fix PR3401: when using large integers, the type by Duncan Sands · 17 years ago
- db7c5f6 Move CurDebugLoc into SelectionDAGLowering. by Dale Johannesen · 17 years ago
- 555a375 Make LowerCallTo and LowerArguments take a DebugLoc by Dale Johannesen · 17 years ago
- ed255b3 Propagate debug info when building SelectionDAG. by Dale Johannesen · 17 years ago
- 4aa1846 Make isOperationLegal do what its name suggests, and introduce a by Dan Gohman · 17 years ago
- ba21b7d Formatting. by Duncan Sands · 17 years ago
- 0bd9546 Delete redundant return statements. by Dan Gohman · 17 years ago
- 4b46224 During bittest switch lowering emit shift in the test block, which should (theoretically) by Anton Korobeynikov · 17 years ago
- be7e414 Cleanup whitespace and comments, and tweak some by Duncan Sands · 17 years ago
- a70798c Allow targets to legalize operations (with illegal operands) that produces multiple values. For example, a load with an illegal operand (a load produces two values, a value and chain). by Sanjiv Gupta · 17 years ago
- 2395916 Use "SINT_TO_FP" instead of "UINT_TO_FP" when getting the exponent. This was by Bill Wendling · 17 years ago
- 786a683 Shift types need to match. by Bill Wendling · 17 years ago
- 44afc82 Verify debug info. by Devang Patel · 17 years ago
- 6e8d814 Registry.h should not depend on CommandLine.h. by Mikhail Glushenkov · 17 years ago
- b2f9a73 Delete trailing whitespace. by Mikhail Glushenkov · 17 years ago
- 76d190c by Devang Patel · 17 years ago
- 6de96a1 Add the private linkage. by Rafael Espindola · 17 years ago
- 91febd1 More consts on TargetLowering references. by Dan Gohman · 17 years ago
- 08a4c28 minor refactoring: use a more specific API by Gabor Greif · 17 years ago
- 5c6e1e3 by Devang Patel · 17 years ago
- dc020f9 Rename getABITypeSize to getTypePaddedSize, as suggested by Chris. by Duncan Sands · 17 years ago
- c7847cd Fix a bug in ComputeLinearIndex computation handling multi-level by Dan Gohman · 17 years ago
- 56a8bb6 by Devang Patel · 17 years ago
- 8feb694 Fix PR3274: when promoting the condition of a BRCOND node, by Duncan Sands · 17 years ago
- f4a66e8 Restore debug printing by Anton Korobeynikov · 17 years ago
- d305d00 Sometimes APInt syntax is really ugly... :( by Anton Korobeynikov · 17 years ago
- 05149ba Indent stuff properly by Anton Korobeynikov · 17 years ago
- 6f21913 Initial checkin of APInt'ififcation of switch lowering by Anton Korobeynikov · 17 years ago
- 12f2490 Clean up the atomic opcodes in SelectionDAG. by Dan Gohman · 17 years ago
- c35fc49 We have decided not to support inline asm where an output operand with a matching input operand with incompatible type (i.e. either one is a floating point and the other is an integer or the sizes of the types differ). SelectionDAGBuild will catch these and exit with an error. by Evan Cheng · 17 years ago
- 4eb2dcd Whitespace fixes. by Bill Wendling · 17 years ago
- db8ec2d Add sub/mul overflow intrinsics. This currently doesn't have a by Bill Wendling · 17 years ago
- 3d96094 There are no longer any places that require a by Duncan Sands · 17 years ago
- 6683547 - Make lowering of "add with overflow" customizable by back-ends. by Bill Wendling · 17 years ago
- 2967480 Added check to avoid generating extract subvector beyond the end of the vector when normalizing vector shuffles. by Mon P Wang · 17 years ago
- dd70834 reapply Sanjiv's patch to genericize memcpy/memset/memmove to take an by Chris Lattner · 17 years ago
- 4bce2bf Revert r59802. It was breaking the build of llvm-gcc: by Bill Wendling · 17 years ago