summaryrefslogtreecommitdiffstats
path: root/llvm/lib
Commit message (Collapse)AuthorAgeFilesLines
* In the below scenario, we must be able to skip the a DBG_VALUE instruction andSumanth Gundapaneni2017-01-091-3/+8
| | | | | | | | | | | | | | | remove the dead store. %vreg0<def> = L2_loadri_io <fi#15>, 0; mem:LD4[%dataF](align=4) DBG_VALUE %vreg0, %noreg, !"dataF", <!184>; IntRegs:%vreg0 S2_storeri_io <fi#15>, 0, %vreg0; mem:ST4[%dataF] In reality, this kind of stores are eliminated before Stack Slot Coloring pass, possibly in instruction lowering Differential Revision: https://reviews.llvm.org/D26616 llvm-svn: 291455
* [X86][AVX512] Enable v16i8/v32i8 vector shifts to use an ↵Simon Pilgrim2017-01-091-14/+17
| | | | | | | | | | extend+shift+truncate pattern. Use the existing AVX2 v8i16 vector shift lowering for v16i8 (extending to v16i32) on AVX512 targets and v32i8 (extending to v32i16) on AVX512BW targets. Cost model updates to follow. llvm-svn: 291451
* fix comment typos; NFCSanjay Patel2017-01-091-5/+5
| | | | llvm-svn: 291447
* [X86][AVX512DQ] Enable v16i16 vector shifts to use an extend+shift+truncate ↵Simon Pilgrim2017-01-091-3/+4
| | | | | | | | | | pattern. Use the existing AVX2 v8i16 vector shift lowering for v16i16 on AVX512 targets (AVX512BW will have already have lowered with vpsravw). Cost model updates to follow. llvm-svn: 291445
* Some formatting in TargetMachineC. NFCAmaury Sechet2017-01-091-2/+2
| | | | llvm-svn: 291442
* [SelectionDAG] Fix in legalization of UMAX/SMAX/UMIN/SMIN. Solves PR31486.Bjorn Pettersson2017-01-091-2/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | Summary: Originally i64 = umax t8, Constant:i64<4> was expanded into i32,i32 = umax Constant:i32<0>, Constant:i32<0> i32,i32 = umax t7, Constant:i32<4> Now instead the two produced umax:es return i32 instead of i32, i32. Thanks to Jan Vesely for help with the test case. Patch by mikael.holmen at ericsson.com Reviewers: bogner, jvesely, tstellarAMD, arsenm Subscribers: test, wdng, RKSimon, arsenm, nhaehnle, llvm-commits Differential Revision: https://reviews.llvm.org/D28135 llvm-svn: 291441
* Fix MSVC build failure introduced in r291431Pavel Labath2017-01-091-4/+3
| | | | | | | MSVC does not like to reinterpret_cast to a uint64_t. Use a different cast instead. llvm-svn: 291435
* RuntimeDyldELF: don't create thunk if not neededEugene Leviant2017-01-092-1/+47
| | | | | | | | | | | | | This patch doesn't create thunk for branch operation when following conditions are met: - Architecture is AArch64 - Relocation target is in the same object file - Relocation target is close enough to be encoded in immediate offset In such case we branch directly to the target instead of branching to thunk Differential revision: https://reviews.llvm.org/D28108 llvm-svn: 291431
* [PM] Teach SCEV to invalidate itself when its dependencies becomeChandler Carruth2017-01-091-0/+12
| | | | | | | | | | | | | invalid. This fixes use-after-free bugs that will arise with any interesting use of SCEV. I've added a dedicated test that works diligently to trigger these kinds of bugs in the new pass manager and also checks for them explicitly as well as triggering ASan failures when things go squirly. llvm-svn: 291426
* [WebAssembly] Fix the opcode values for i64.eq and i64.ne.Dan Gohman2017-01-091-2/+2
| | | | llvm-svn: 291424
* Remove unused method in LoopVectorize.cpp.Jonas Paulsson2017-01-091-7/+0
| | | | | | | computeInterleaveCount() is not defined/used and is therefore removed. Review: Davide Italiano llvm-svn: 291423
* NewGVN: Fix PR 31573, a failure to verify memory congruency due toDaniel Berlin2017-01-091-1/+14
| | | | | | | not excluding ourselves when checking if any equivalent stores exist. llvm-svn: 291421
* NewGVN: Change a std::vector to SmallVector and cleanup naming.Daniel Berlin2017-01-091-10/+11
| | | | llvm-svn: 291420
* [AVX-512] Change another pattern that was using BLENDM to use masked moves. ↵Craig Topper2017-01-091-15/+24
| | | | | | A future patch will conver it back to BLENDM if its beneficial to register allocation. llvm-svn: 291419
* [AVX-512] Add patterns to use a zero masked VPTERNLOG instruction for ↵Craig Topper2017-01-092-0/+31
| | | | | | | | vselects of all ones and all zeros. Previously we emitted a VPTERNLOG and a separate masked move. llvm-svn: 291415
* Define sys::path::convert_to_slashRui Ueyama2017-01-092-10/+12
| | | | | | | | This patch moves convertToUnixPathSeparator from LLD to LLVM. Differential Revision: https://reviews.llvm.org/D28444 llvm-svn: 291414
* CommandLine option: Relax the assertion introduced in r290467 to allows for ↵Mehdi Amini2017-01-081-1/+1
| | | | | | | | | empty string This is used in LDC for custom boolean commandline options, setArgStr is called with an empty string before using AddLiteralOption. llvm-svn: 291406
* [MemDep] NFC walk invariant.group graph only downPiotr Padlewski2017-01-081-26/+16
| | | | | | | | | | | | | | Summary: By using stripPointerCasts we can get to the root value and then walk down the bitcast graph Reviewers: reames Subscribers: llvm-commits Differential Revision: https://reviews.llvm.org/D28181 llvm-svn: 291405
* [LCSSA] Fix some typos. NFCI.Davide Italiano2017-01-081-3/+3
| | | | llvm-svn: 291404
* [AVX-512] If avx512dq is available use vpmovm2d/vpmovm2q instead of vselect ↵Craig Topper2017-01-081-9/+13
| | | | | | of zeroes/ones when handling sign extends of i1 without VLX. llvm-svn: 291402
* [SCCP] Unknown instructions are sent to overdefined anyway. NFCI.Davide Italiano2017-01-081-18/+0
| | | | llvm-svn: 291400
* llvm-objdump: speed up -objc-meta-dataSaleem Abdulrasool2017-01-081-0/+8
| | | | | | | | | | | | | | | | | | Running a Debug build of objdump -objc-meta-data with a large Mach-O file is currently unnecessarily slow. With some local test input, this change reduces the run time from 75-85s down to 15-20s. The two changes are: Assert on pointer equality not array equality Replace vector<pair<address, symbol>> with DenseMap<address, symbol> Additionally, use a std::unique_ptr rather than handling the memory manually. Patch by Dave Lee! llvm-svn: 291398
* Strip trailing whitespace.Simon Pilgrim2017-01-081-1/+1
| | | | llvm-svn: 291395
* Fix line endings and strip trailing whitespace.Simon Pilgrim2017-01-081-71/+71
| | | | llvm-svn: 291393
* [x86] fix usage of stale operands when lowering selectSanjay Patel2017-01-081-2/+9
| | | | | | | | | | | | | | | | | | | | | I noticed this problem as part of the ongoing attempt to canonicalize min/max ops in IR. The debug output shows nodes like this: t4: i32 = xor t2, Constant:i32<-1> t21: i8 = setcc t4, Constant:i32<0>, setlt:ch t14: i32 = select t21, t4, Constant:i32<-1> And because the select is holding onto the t4 (xor) node while EmitTest creates a new x86-specific xor node, the lowering results in: t4: i32 = xor t2, Constant:i32<-1> t25: i32,i32 = X86ISD::XOR t2, Constant:i32<-1> t28: i32,glue = X86ISD::CMOV Constant:i32<-1>, t4, Constant:i8<15>, t25:1 Differential Revision: https://reviews.llvm.org/D28374 llvm-svn: 291392
* [CostModel][X86] Fixed vXi8 uniform shift costs.Simon Pilgrim2017-01-081-6/+16
| | | | | | | | | | The 'fast' costs should only work for shifts by uniform constants (uniform non-constant are lowered using the slow default implementation). Logical shifts were not taking into account that we must mask the psrlw result, so the costs needed to be doubled. Added missing AVX2/AVX512BW costs as well. llvm-svn: 291391
* [CostModel][X86] Moved legal uniform shift costs earlier.Simon Pilgrim2017-01-081-24/+39
| | | | | | XOP was prematurely matching, doubling the cost of ashr/lshr uniform shifts. llvm-svn: 291390
* [AVX-512] Remove redundant patterns that select unaligned moves with zero ↵Craig Topper2017-01-081-1/+1
| | | | | | masking for patterns that already use the aligned form. NFC llvm-svn: 291383
* [ThinLTO] Fix lazy-loading of Metadata attachment, which left some Fwd ref ↵Mehdi Amini2017-01-081-1/+2
| | | | | | | | | | | behind The change in r291362 was too agressive. We still need to flush at the end of the block because function local metadata can introduce fwd ref as well. (Bootstrap with ThinLTO was broken) llvm-svn: 291379
* [ThinLTO] Expected<> return values need to be handled to avoid an assertionMehdi Amini2017-01-081-1/+8
| | | | llvm-svn: 291377
* [AVR] Implement TargetLoweing::getRegisterByNameDylan McKay2017-01-072-0/+44
| | | | | | | This allows the use of the 'read_register' intrinsics used by clang's named register globals features. llvm-svn: 291375
* [CostModel][X86] Update SSE41/AVX1 vXi32 SHL costsSimon Pilgrim2017-01-071-0/+2
| | | | | | SSE41 provides pmulld which allows the simpler pslld/paddd/cvttps2dq/pmulld pattern than SSE2's use of pmuludq. llvm-svn: 291372
* [AVX-512] Remove patterns from the other VBLENDM instructions. They are all ↵Craig Topper2017-01-071-12/+5
| | | | | | | | redundant with masked move instructions. We should probably teach the two address instruction pass to turn masked moves into BLENDM when its beneficial to the register allocator. llvm-svn: 291371
* [AVX-512] Remove patterns from masked broadcast versions of BLENDM instructions.Craig Topper2017-01-071-6/+3
| | | | | | | | All but (v2f64 broadcast f64) are handled with VBROADCAST instructions. The v2f64 version can be handled with VMOVDDUP. We may want to consider converting to BLENDM instructions in the two address instruction pass if its beneficial to register allocation. llvm-svn: 291369
* [AVX-512] Add masked forms of the alternate MOVDDUP patterns.Craig Topper2017-01-071-0/+22
| | | | | | I'm not too sure how to get isel to select even all of the unmasked forms, but at least we have a consistent set now. llvm-svn: 291368
* [CostModel][X86] Fix AVX2 v16i16 shift 'splat' costs.Simon Pilgrim2017-01-071-2/+15
| | | | llvm-svn: 291366
* [CostModel][X86] Match 256-bit vector shift 'splat' costs for AVX2 and aboveSimon Pilgrim2017-01-071-45/+44
| | | | | | We were matching against general vector shift costs before the uniform splat costs llvm-svn: 291365
* [CostModel][X86] Generalized cost calculation of SHL by constant -> MUL ↵Simon Pilgrim2017-01-071-21/+10
| | | | | | conversion. llvm-svn: 291364
* [ThinLTO] Fix assertions on lazy-loading of Metadata TBAA attachmentsMehdi Amini2017-01-071-4/+4
| | | | | | | | | | | | | | | | | | | | | Summary: The issue happens with: %0 = ....., !tbaa !0 %1 = ....., !tbaa !1 With !0 that references !1. In this case when loading !0 we generates a temporary for the operand !1. We now flush it immediately and trigger the load of !1 before moving on. If we don't we get the temporary when attaching to %1. This is usually not an issue except that we eagerly try to update TBAA MDNodes, which is obviously not possible if we only have a temporary. Differential Revision: https://reviews.llvm.org/D28423 llvm-svn: 291362
* SimplifyLibCalls: Remove incorrect optimization of fabsMatt Arsenault2017-01-071-10/+2
| | | | | | | | fabs(x * x) is not generally safe to assume x is positive if x is a NaN. This is also less general than it could be, so this will be replaced with a transformation on the intrinsic. llvm-svn: 291359
* [Bitcode] Remove unused PlaceHolder parameter to lazyLoadModuleMetadataBlock()Mehdi Amini2017-01-071-4/+4
| | | | llvm-svn: 291356
* [CostModel][X86] Merge separate AVX1 cost LUTs. NFCI.Simon Pilgrim2017-01-071-38/+30
| | | | llvm-svn: 291355
* [CostModel][AVX512BW] Add v32i16 vector shift costs for avx512bw targets.Simon Pilgrim2017-01-071-0/+4
| | | | llvm-svn: 291354
* [CostModel][X86] Added missing AVX2 arithmetic costs.Simon Pilgrim2017-01-071-23/+33
| | | | | | Allows us to correctly fall through to the lower AVX1 costs if look up failed. llvm-svn: 291353
* [CostModel][X86] Reordered AVX1 arithmetic cost LUT into descending target ↵Simon Pilgrim2017-01-071-27/+27
| | | | | | order. NFCI. llvm-svn: 291352
* NewGVN: Make sure we properly lookup operand leaders while creatingDaniel Berlin2017-01-071-13/+48
| | | | | | | congruence classes for stores, and then keep them up to date. Add testcases. llvm-svn: 291351
* [X86][AVX512] Use lowerShuffleAsRepeatedMaskAndLanePermute for non-VBMI ↵Simon Pilgrim2017-01-072-2/+7
| | | | | | v64i8 shuffles (PR31470) llvm-svn: 291347
* TarWriter: Use fitsInUstar function.Rui Ueyama2017-01-071-2/+2
| | | | | | This change should have been commit as part of r291340. llvm-svn: 291341
* TarWriter: Use Ustar header's "prefix" field to store long filenames.Rui Ueyama2017-01-071-1/+27
| | | | | | | | | | | | Tar's Ustar header has the "prefix" field to store a directory part of a filename. It is not as flexible as the PAX-extended filename because there's still a limitation on the maximum filename size, but it mitigates the situation. This patch should unbreak some Windows buildbots that uses very old tar command. llvm-svn: 291340
* [X86] Disable load unfolding for 128-bit MOVDDUP instructions since the load ↵Craig Topper2017-01-071-2/+2
| | | | | | size is smaller than the register size so unfolding would increase the load size. llvm-svn: 291338
OpenPOWER on IntegriCloud