- ed5707b DAGCombine should not aggressively fold SEXT(VSETCC(...)) into a wider VSETCC without first checking the target's vector boolean contents. by Owen Anderson · 12 years ago
- 0cb1019 Legalize vector truncates by parts rather than just splitting. by Jim Grosbach · 12 years ago
- 1e48093 Tidy up comment grammar. by Jim Grosbach · 12 years ago
- 6265d5c Remove unused MEMBARRIER DAG node; it's been replaced by ATOMIC_FENCE. by Tim Northover · 12 years ago
- 456ca04 Add CodeGen support for functions that always return arguments via a new parameter attribute 'returned', which is taken advantage of in target-independent tail call opportunity detection and in ARM call lowering (when placed on an integral first parameter). by Stephen Lin · 12 years ago
- 462123f Simplify the code in FastISel::tryToFoldLoad, add an assertion and fix a comment. by Eli Bendersky · 12 years ago
- 75299e3 Move TryToFoldFastISelLoad to FastISel, where it belongs. In general, I'm by Eli Bendersky · 12 years ago
- 2a8bea7 ArrayRefize getMachineNode(). No functionality change. by Michael Liao · 12 years ago
- 03494e0 Use dbgs() consistently for -debug printouts by Eli Bendersky · 12 years ago
- 5bd0767 Add some more stats for fast isel vs. SelectionDAG, w.r.t lowering function by Eli Bendersky · 12 years ago
- 6fac1fb DAGCombiner: Fold a shuffle on CONCAT_VECTORS into a new CONCAT_VECTORS if possible. by Benjamin Kramer · 12 years ago
- 5622eae Formatting by Eli Bendersky · 12 years ago
- cd7a155 Fix PR15632: No support for ppcf128 floating-point remainder on PowerPC. by Bill Schmidt · 12 years ago
- e737018 DAGCombiner: Merge store/loads when we have extload/truncstores by Arnold Schwaighofer · 12 years ago
- f28a29b Merge load/store sequences with adresses: base + index + offset by Arnold Schwaighofer · 12 years ago
- 0b68b75 DAGCombine: visitXOR can replace a node without returning it, bail out in that case. by Benjamin Kramer · 12 years ago
- 7ae3bb8 [fast-isel] Add a preemptive fix for the case where we fail to materialize an by Chad Rosier · 12 years ago
- b4f98ea Enhance folding of (extract_subvec (insert_subvec V1, V2, IIdx), EIdx) by Michael Liao · 12 years ago
- 1cd1d02 Disable some unsafe-fp-math DAG-combine transformation after legalization. by Shuxin Yang · 12 years ago
- 7834c71 Remove the type legality check from the SelectionDAGBuilder when it lowers @llvm.fmuladd to ISD::FMA nodes. by Owen Anderson · 12 years ago
- 93c1fd4 Make variable name more explicit and eliminate redundant lookup in SDNodeOrdering by Justin Holewinski · 12 years ago
- 7745175 When computing the demanded bits of Load SDNodes, make sure that we are looking at the loaded-value operand and not the ptr result (in case of pre-inc loads). by Nadav Rotem · 12 years ago
- 61aba54 Revert "pre-RA-sched: fix TargetOpcode usage" by Christian Konig · 12 years ago
- fa963a8 Move SDNode order propagation to SDNodeOrdering, which also fixes a missed by Justin Holewinski · 12 years ago
- 0609151 pre-RA-sched: fix TargetOpcode usage by Christian Konig · 12 years ago
- d73dc54 Propagate DAG node ordering during type legalization and instruction selection by Justin Holewinski · 12 years ago
- 4cb1f5f Reset some of the target options which affect code generation. by Bill Wendling · 12 years ago
- 1a5c051 Avoid generating ISD::SELECT for vector operands to SIGN_EXTEND by Richard Relph · 12 years ago
- 33cdfe9 Fix a crasher newly introduced in r176659/r176649, where fast-isel tries to by Nick Lewycky · 12 years ago
- fa785cb Disable statistics on Release builds and move tests that depend on -stats. by Jan Wen Voung · 12 years ago
- 7a58099 Remove default from fully covered switch. by Benjamin Kramer · 12 years ago
- 03abf2f LegalizeDAG: Respect the result of TLI.getBooleanContents() when expanding SETCC by Tom Stellard · 12 years ago
- 3ef5383 DAGCombiner: Use correct value type for checking legality of BR_CC v3 by Tom Stellard · 12 years ago
- e36b47e Revert r176154 in favor of a better approach. by Bill Wendling · 12 years ago
- 4fde76d [fast-isel] Seriously, add support for the expect intrinsic. rdar://13370942 by Chad Rosier · 12 years ago
- 33947b4 [fast-isel] Add support for the expect intrinsic. rdar://13370942 by Chad Rosier · 12 years ago
- 3853f74 ArrayRefize some code. No functionality change. by Benjamin Kramer · 12 years ago
- 877fcf5 pre-RA-sched debug-only fix by Andrew Trick · 12 years ago
- ccd38f4 pre-RA-sched assertion fix. This bug was exposed by r176037. by Andrew Trick · 12 years ago
- 4d89545 No need to go through int64 and APInt when generating a new constant. by Nadav Rotem · 12 years ago
- 34fd0d2 SDAG: Handle scalarizing an extend of a <1 x iN> vector. by Jim Grosbach · 12 years ago
- b804a1b Remove duplicate line and move another closer to its actual use by Eli Bendersky · 12 years ago
- c3c3810 Set properties for f128 type. by Akira Hatanaka · 12 years ago
- 7590022 Generate an error message instead of asserting or segfaulting when we can't by Chad Rosier · 12 years ago
- a6b20ce Fix PR10475 by Michael Liao · 12 years ago
- 6437d38 A small refactoring + adding comments. by Eli Bendersky · 12 years ago
- 0a4da5f Put some per-instruction statistics of fast isel under NDEBUG, together with by Eli Bendersky · 12 years ago
- f686cfb Remove unnecessary cast to void. by Eric Christopher · 12 years ago
- 139f50a Silence the unused variable warning. by Nadav Rotem · 12 years ago
- 07cd932 The FastISEL should be fast. But when we record statistics we use atomic operations to increment the counters. by Nadav Rotem · 12 years ago
- 7dbd34b Reverted: r176136 - Have a way for a target to opt-out of target-independent fast isel by Michael Ilseman · 12 years ago
- 5e5974f SelectionDAG: If llvm.donothing has a landingpad, we should clear by Manman Ren · 12 years ago
- bdc98d3 Have a way for a target to opt-out of target-independent fast isel by Michael Ilseman · 12 years ago
- ffa1dba Fix wording. by Chad Rosier · 12 years ago
- fd3417d [fast-isel] Add X86FastIsel::FastLowerArguments to handle functions with 6 or by Chad Rosier · 12 years ago
- 029f4fd pre-RA-sched fix: only reevaluate physreg interferences when necessary. by Andrew Trick · 12 years ago
- 50e75bf 'Hexadecimal' has two 'a's and only one 'i'. by Matt Beaumont-Gay · 12 years ago
- 7e6ffac Fix spelling noticed by Duncan. by Chandler Carruth · 12 years ago
- af23f8e Fix the root cause of PR15348 by correctly handling alignment 0 on by Chandler Carruth · 12 years ago
- d99a5a3 SelectionDAG compile time improvement. by Nadav Rotem · 12 years ago
- b704ffb Fix isa<> check which could never be true. by Pete Cooper · 12 years ago
- f31bd0f DAGCombiner: Make the post-legalize vector op optimization more aggressive. by Benjamin Kramer · 12 years ago
- c46e2df DAGCombiner: Fold pointless truncate, bitcast, buildvector series by Arnold Schwaighofer · 12 years ago
- eedff35 Fix PR15267 by Michael Liao · 12 years ago
- ad4da0f Move the SplatByte helper to APInt and generalize it a bit. by Benjamin Kramer · 12 years ago
- 751bc8d Fix #includes, so we include only what we really need. by Jakub Staszak · 12 years ago
- b875acd [ms-inline asm] Remove a redundant call to the setHasMSInlineAsm function. by Chad Rosier · 12 years ago
- 3ef669c [ms-inline asm] Fix undefined behavior to reset hasMSInlineAsm in advance of SelectAllBasicBlocks(). by NAKAMURA Takumi · 12 years ago
- 631dd86 LegalizeDAG.cpp doesn't need DenseMap. by Jakub Staszak · 12 years ago
- b566062 [ms-inline asm] Do not omit the frame pointer if we have ms-inline assembly. by Chad Rosier · 12 years ago
- 789cb5d Use the 'target-features' and 'target-cpu' attributes to reset the subtarget features. by Bill Wendling · 12 years ago
- 86cdbc9 enable SDISel sincos optimization for GNU environments by Paul Redmond · 12 years ago
- 6cc4b8d Dont merge consecutive loads/stores into vectors when noimplicitfloat is used. by Nadav Rotem · 12 years ago
- 39125d9 Add some legality checks for SETCC before introducing it in the DAG combiner post-operand legalization. by Owen Anderson · 12 years ago
- 87d0b9e Add static cast to unsigned char whenever a character classification function is called with a signed char argument, in order to avoid assertions in Windows Debug configuration. by Guy Benyei · 12 years ago
- 5c97450 PR14562 - Truncation of left shift became undef by Paul Redmond · 12 years ago
- b0fde6d Check type for legality before forming a select from loads. by Pete Cooper · 12 years ago
- 092e5e7 Currently, codegen may spent some time in SDISel passes even if an entire by Evan Cheng · 12 years ago
- 5f352cc Remove unnecessary code. by Evan Cheng · 12 years ago
- 089a5f8 DAGCombiner: Constant folding around pre-increment loads/stores by Hal Finkel · 12 years ago
- 8f637ad Revert 172027 and 174336. Remove diagnostics about over-aligned stack objects. by Bob Wilson · 12 years ago
- 607acd6 Tweak check to avoid integer overflow (for insanely large alignments) by Evan Cheng · 12 years ago
- b48783b Reapply r174343, with a fix for a scary DAG combine bug where it failed to differentiate between the alignment of the by Owen Anderson · 12 years ago
- eb260b2 Revert r174343, "When the target-independent DAGCombiner inferred a higher alignment for a load," by NAKAMURA Takumi · 12 years ago
- 429f7ef When the target-independent DAGCombiner inferred a higher alignment for a load, by Owen Anderson · 12 years ago
- 4969310 SelectionDAG: Teach FoldConstantArithmetic how to deal with vectors. by Benjamin Kramer · 12 years ago
- 98b93e5 rdar://13126763 by Shuxin Yang · 12 years ago
- 5882e56 Fix errant fallthrough in the generation of the lifetime markers. by Nadav Rotem · 12 years ago
- 2d95e43 When lowering memcpys to loads and stores, make sure we don't promote alignments by Lang Hames · 12 years ago
- f662a59 Add a special handling case for untyped CopyFromReg node in GetCostForDef() of ScheduleDAGRRList by Weiming Zhao · 12 years ago
- 8688a58 Teach SDISel to combine fsin / fcos into a fsincos node if the following by Evan Cheng · 12 years ago
- 0fef46f Legalizer: Reword comment again, per Duncan's suggestion. by Benjamin Kramer · 12 years ago
- 022688c Legalizer: Add an assert and tweak a comment to clarify the assumptions this code makes. by Benjamin Kramer · 12 years ago
- 914f8c4 When the legalizer is splitting vector shifts, the result may not have the right shift amount type. by Benjamin Kramer · 12 years ago
- ea387fc This patch aims to reduce compile time in LegalizeTypes by using SmallDenseMap, by Preston Gurd · 12 years ago
- 0a29cb0 Make APFloat constructor require explicit semantics. by Tim Northover · 12 years ago
- 0c8607b by Nadav Rotem · 12 years ago
- 1b0c54f Use AttributeSet accessor methods instead of Attribute accessor methods. by Bill Wendling · 12 years ago
- 1a17bd2 Remove unused parameter. Also use the AttributeSet query methods instead of the Attribute query methods. by Bill Wendling · 12 years ago
- 6c327f9 Optimization for the following SIGN_EXTEND pairs: by Elena Demikhovsky · 12 years ago