summaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
...
* [PGO] Profile summary reader/writer supportXinliang David Li2016-02-038-22/+259
| | | | | | | | | | With this patch, the profile summary data will be available in indexed profile data file so that profiler reader/compiler optimizer can start to make use of. Differential Revision: http://reviews.llvm.org/D16258 llvm-svn: 259626
* LowerBitSets: Don't bother to do any work if the llvm.bitset.test intrinsic ↵Peter Collingbourne2016-02-031-1/+1
| | | | | | is unused. llvm-svn: 259625
* Make CF constant string decl visible to name lookup to fix module errorsBen Langmuir2016-02-0311-18/+54
| | | | | | | | | | | | | | | | | | | | | | | | | | | The return type of the __builtin___*StringMakeConstantString functions is a pointer to a struct, so we need that struct to be visible to name lookup so that we will correctly merge multiple declarations of that type if they come from different modules. Incidentally, to make this visible to name lookup we need to rename the type to __NSConstantString, since the real NSConstantString is an Objective-C interface type. This shouldn't affect anyone outside the compiler since users of the constant string builtins cast the result immediately to CFStringRef. Since this struct type is otherwise implicitly created by the AST context and cannot access namelookup, we make this a predefined type and initialize it in Sema. Note: this issue of builtins that refer to types not visible to name lookup technically also affects other builtins (e.g. objc_msgSendSuper), but in all other cases the builtin is a library builtin and the issue goes away if you include the library that defines the types it uses, unlike for these constant string builtins. rdar://problem/24425801 llvm-svn: 259624
* Add #include "llvm/Support/raw_ostream.h" to fix Windows build.Peter Collingbourne2016-02-031-0/+1
| | | | llvm-svn: 259623
* Fix miscompile and rejects-valids when disambiguating after an ambiguousRichard Smith2016-02-032-5/+19
| | | | | | | C-style-cast to function/array type or parenthesized function-style cast/array indexing. llvm-svn: 259622
* Transforms: Move GlobalOpt's Evaluator to Utils where it can be reused.Peter Collingbourne2016-02-034-655/+717
| | | | llvm-svn: 259621
* Fix typo in comment. NFCNick Lewycky2016-02-031-1/+1
| | | | llvm-svn: 259620
* docs: Document how bitsets may be used to encode type information.Peter Collingbourne2016-02-031-11/+117
| | | | llvm-svn: 259619
* Fix typo in OpenCL type mangling. This is still bogus (we should either use theRichard Smith2016-02-031-1/+1
| | | | | | | | | | actual source name of the typedef or a vendor mangling) but at least it stands a chance of demangling now. We would also benefit from some test coverage of this (OpenCL + __attribute__((overloadable)) is probably required to reach this). llvm-svn: 259618
* Codegen: [PPC] Fix PPCVSXFMAMutate to handle duplicates.Kyle Butt2016-02-032-19/+68
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | The purpose of PPCVSXFMAMutate is to elide copies by changing FMA forms on PPC. %vreg6<def> = COPY %vreg96 %vreg6<def,tied1> = XSMADDASP %vreg6<tied0>, %vreg5<kill>, %vreg7 ;v6 = v6 + v5 * v7 is replaced by %vreg5<def,tied1> = XSMADDMSP %vreg5<tied0>, %vreg7, %vreg96 ;v5 = v5 * v7 + v96 This was broken in the case where the target register was also used as a multiplicand. Fix this case by checking for it and replacing both uses with the copied register. %vreg6<def> = COPY %vreg96 %vreg6<def,tied1> = XSMADDASP %vreg6<tied0>, %vreg5<kill>, %vreg6 ;v6 = v6 + v5 * v6 is replaced by %vreg5<def,tied1> = XSMADDMSP %vreg5<tied0>, %vreg96, %vreg96 ;v5 = v5 * v96 + v96 llvm-svn: 259617
* Fix Itanium RTTI emission so that we emit fundamental type information into theRichard Smith2016-02-032-2/+21
| | | | | | | | | | | | | | | | C++ ABI library for the same set of types for which we expect the C++ ABI library to provide the RTTI. Specifically: 1) __int128 and unsigned __int128 are now emitted into the ABI library. We always expected them to be there but never actually made sure to emit them. 2) Do not expect OpenCL builtin types to have type info in the C++ ABI library. Neither libc++abi nor libstdc++ puts them there when built with either GCC or Clang. This matches GCC's behavior. llvm-svn: 259616
* Revert r259576: Disable the vzeroupper insertion pass on PS4.Yunzhong Gao2016-02-032-12/+0
| | | | | | Will re-implement based on review feedback. llvm-svn: 259615
* RegCoalescer: Making sure re-materialization defines all subrangesMarcello Maggioni2016-02-031-0/+30
| | | | | | | | | | | | | The register coalescer can rematerialize constants that define more of a register than the copy it is going to replace was going to do. This is valid in the case the register was undef before the copy happened. This patch makes sure that all the subranges defined by the new rematerialization instructions have at least a dead def. Review: http://reviews.llvm.org/D16693 llvm-svn: 259614
* Define SymbolBody::getSize instead of getSymSize(SymbolBody&). NFC.Rui Ueyama2016-02-033-9/+15
| | | | llvm-svn: 259613
* DiagnosticInfoWithDebugLocBase: Appease Twine for now.NAKAMURA Takumi2016-02-031-2/+2
| | | | | FIXME: We should get rid of Twine in the record. llvm-svn: 259612
* The compiler may use "line number 0" to indicate compiler generated goo that ↵Jim Ingham2016-02-032-2/+50
| | | | | | | | | | | it can't track a source for. When we are pushing breakpoints and stepping past function prologues, also push past code from line 0 immediately following the prologue end. <rdar://problem/23730696> llvm-svn: 259611
* [LoopVersioning] Expose loop versioning as a pass tooAdam Nemet2016-02-035-0/+129
| | | | | | | | | | | | | | | | | | | | | | | | | | | Summary: LoopVersioning is a transform utility that transform passes can use to run-time disambiguate may-aliasing accesses. I'd like to also expose as pass to allow it to be unit-tested. I am planning to add support for non-aliasing annotation in LoopVersioning and I'd like to be able to write tests directly using this pass. (After that feature is done, the pass could also be used to look for optimization opportunities that are hidden behind incomplete alias information at compile time.) The pass drives LoopVersioning in its default way which is to fully disambiguate may-aliasing accesses no matter how many checks are required. Reviewers: hfinkel, ashutosh.nema, sbaranga Subscribers: zzheng, mssimpso, llvm-commits, sanjoy Differential Revision: http://reviews.llvm.org/D16612 llvm-svn: 259610
* PR24989: Stop trying to use the C++11 rules for lambda return type inference inRichard Smith2016-02-023-14/+26
| | | | | | | | C++14 generic lambdas. It conflicts with the C++14 return type deduction mechanism, and results in us failing to actually deduce the lambda's return type in some cases. llvm-svn: 259609
* Fix inverted conditional in TestInferiorAssert.pyZachary Turner2016-02-021-2/+2
| | | | llvm-svn: 259608
* Fix a thinko in StackFrame::GetInScopeVariableList.Siva Chandra2016-02-021-1/+1
| | | | | | | | | | Reviewers: clayborg Subscribers: lldb-commits Differential Revision: http://reviews.llvm.org/D16745 llvm-svn: 259607
* ELF: Write about the current policy to README.Rui Ueyama2016-02-021-0/+13
| | | | llvm-svn: 259606
* Set correct thread stop info when single-step lands on a breakpoint [Windows]Adrian McCarthy2016-02-021-5/+19
| | | | | | | | I don't understand how this worked before, but this fixes the recent test regressions on Windows in TestConsecutiveBreakpoints.py. Differential Revision: http://reviews.llvm.org/D16825 llvm-svn: 259605
* Fix rejects-valid when forming a pointer-to-member with 'decltype(expr)::*'.Richard Smith2016-02-022-1/+3
| | | | llvm-svn: 259604
* Attempt #2 to unbreak r259595.George Burgess IV2016-02-021-4/+4
| | | | llvm-svn: 259602
* [codeview] Improve readability of codeview assembly outputDavid Majnemer2016-02-026-109/+200
| | | | | | | Strictly speaking, this is not an improvement in functionality per se but a usability improvement to those debugging codeview. llvm-svn: 259601
* [libFuzzer] don't create too many trace-based mutations as it may be too slowKostya Serebryany2016-02-021-2/+18
| | | | llvm-svn: 259600
* Attempt to fix builds broken by r259595.George Burgess IV2016-02-022-2/+2
| | | | llvm-svn: 259599
* Work around build failure due to GCC 4.8.1 bug. We don't completely understandRichard Smith2016-02-021-14/+6
| | | | | | | | | | the details of the bug, but avoiding overloading llvm::cast with another function template sidesteps it. See gcc.gnu.org/PR58022 for details of the bug, and llvm.org/PR26362 for more backgound on how it manifested in Clang. Patch by Igor Sugak! llvm-svn: 259598
* ELF: Make link() to take an output stream to which error messages are written.Rui Ueyama2016-02-026-5/+13
| | | | | | http://reviews.llvm.org/D16668 llvm-svn: 259597
* ELF: Do not exit if it cannot open an output file.Rui Ueyama2016-02-022-5/+20
| | | | | | | | It can fail to open an output file for various reasons, including lack of permission, too long filename, or the output file is not a mmap'able file. llvm-svn: 259596
* This patch adds MemorySSA to LLVM.George Burgess IV2016-02-0220-0/+2358
| | | | | | | | | Please see include/llvm/Transforms/Utils/MemorySSA.h for a description of MemorySSA, and what it does. Differential Revision: http://reviews.llvm.org/D7864 llvm-svn: 259595
* [LVI] Fix debug outputPhilip Reames2016-02-021-3/+3
| | | | | | Due to staleness in a patch I committed yesterday, the debug output was reporting overdefined cases as being undefined. Confusing to say the least. The mistake appears to have only effected the debug output thankfully. llvm-svn: 259594
* [TSan] Use darwin_filter_host_arch to restrict set of test arch on Mac OS.Alexey Samsonov2016-02-022-4/+6
| | | | | | This also reverts r259577 which was a quick-fix to fix buildbots. llvm-svn: 259593
* [CUDA] Do not allow dynamic initialization of global device side variables.Artem Belevich2016-02-027-11/+486
| | | | | | | | | | | | | | In general CUDA does not allow dynamic initialization of global device-side variables. One exception is that CUDA allows records with empty constructors as described in section E2.2.1 of CUDA 7.5 Programming guide. This patch applies initializer checks for all device-side variables. Empty constructors are accepted, but no code is generated for them. Differential Revision: http://reviews.llvm.org/D15305 llvm-svn: 259592
* ObjCXX: fix a crash during typo correction.Manman Ren2016-02-022-2/+34
| | | | | | | | | | | | | For ObjCXX, we can create a CastExpr with Kind being CK_UserDefinedConversion and SubExpr being BlockExpr. Specifically one can return BlockExpr from BuildCXXMemberCallExpr and the result can be used to build a CastExpr. Fix the assumption in CastExpr::getSubExprAsWritten that SubExpr can only be CXXMemberCallExpr. rdar://problem/24364077 llvm-svn: 259591
* Don't return a tuple from the skip test function.Zachary Turner2016-02-023-34/+36
| | | | | | | | | | | Previously we were returning a tuple of (bool, skip_reason) from the tuple function. This makes for some awkward code, especially since a value of True for the first argument implies that the second argument is None, and a value of False implies that the second argument is not None. So it was basically redundant, and with this patch we simply return the skip reason or None directly. llvm-svn: 259590
* __mh_execute_header atoms should be global and never dead strippped.Pete Cooper2016-02-023-2/+9
| | | | | | | | | | | | | | In r259574 I fixed some of the issues with the mach header symbols and DSO handles. This is the next issue whereby the __mh_execute_header has to not be dead stripped, and (to match ld64) should be dynamically referenced. The test here should also have been added in r259574 to make sure that we emit this symbol. But checking that it is not only emitted but also has the correct reference type is fine. llvm-svn: 259589
* [asan] Remove redundant elifAnna Zaks2016-02-021-2/+0
| | | | | | This is a fixup to r259451. llvm-svn: 259588
* Support loads with differently sized types from a single arrayTobias Grosser2016-02-0211-129/+320
| | | | | | | | | | | | | | | | | | | | | | | | | | | | We support now code such as: void multiple_types(char *Short, char *Float, char *Double) { for (long i = 0; i < 100; i++) { Short[i] = *(short *)&Short[2 * i]; Float[i] = *(float *)&Float[4 * i]; Double[i] = *(double *)&Double[8 * i]; } } To support such code we use as element type of the modeled array the smallest element type of all original array accesses. Accesses with larger types are modeled as multiple accesses with the smaller type. For example the second load access is modeled as: { Stmt_bb2[i0] -> MemRef_Float[o0] : 4i0 <= o0 <= 3 + 4i0 } To support jscop-rewritable memory accesses we need each statement instance to only be assigned a single memory location, which will be the address at which we load the value. Currently we obtain this address by taking the lexmin of the access function. We may consider keeping track of the memory location more explicitly in the future. llvm-svn: 259587
* [asan] Add iOS support to AddressSanitzierAnna Zaks2016-02-021-3/+11
| | | | | | Differential Revision: http://reviews.llvm.org/D15625 llvm-svn: 259586
* [LVI] Code motion only [NFC]Philip Reames2016-02-021-64/+62
| | | | | | I introduced a declaration in 259583 to keep the diff readable. This change just moves the definition up to remove the declaration again. llvm-svn: 259585
* Add "REQUIRES: shell" to fix Windows buildbot.Rui Ueyama2016-02-021-0/+9
| | | | llvm-svn: 259584
* [LVI] Refactor to use newly introduced intersect utility Philip Reames2016-02-021-32/+19
| | | | | | | | This patch uses the newly introduced 'intersect' utility (from 259461: [LVI] Introduce an intersect operation on lattice values) to simplify existing code in LVI. While not introducing any new concepts, this change is probably not NFC. The common 'intersect' function is more powerful that the ad-hoc implementations we'd had in a couple of places. Given that, we may see optimizations triggering a bit more often. llvm-svn: 259583
* Remove utils/builditJustin Bogner2016-02-022-493/+0
| | | | | | | The autoconf build system was removed - this doesn't even work and doesn't need to be here. llvm-svn: 259582
* Correct size calculations for ELF filesHemant Kulkarni2016-02-023-0/+46
| | | | llvm-svn: 259578
* [tsan] Disable x86_64h build that was enabled as part of r259542.Daniel Sanders2016-02-021-0/+3
| | | | | | It fails almost every test on clang-stage1-cmake-RA_check. llvm-svn: 259577
* Disable the vzeroupper insertion pass on PS4.Yunzhong Gao2016-02-022-0/+12
| | | | | | | | See comments in test/CodeGen/X86/avx-vzeroupper.ll for more explanation. Original patch by: Sean Silva llvm-svn: 259576
* [Orc] Stub addresses should be based on stub size, not pointer size.Lang Hames2016-02-021-1/+1
| | | | | | | | This didn't affect X86_64, which is the only client of this code at the moment, as stubs and pointers are both 8-bytes there. It will affect other platforms though. llvm-svn: 259575
* Fix handling of mach header and DSO handle symbols.Pete Cooper2016-02-028-40/+62
| | | | | | | | | | | | | | | | | The magic file which contained these symbols inherited from archive which meant that the resolver didn't add the required atoms as archive members only get added when referenced. Instead we now inherit from SimpleFile which always links in the atoms needed. The second issue was in the handling of these symbols when we emit the MachO. The mach header symbol needs to be in the atom list as it gets an offset (0), and being in the atom list makes sure it is emitted to the symbol table. DSO handles are not emitted to the symbol table. rdar://problem/24450654 llvm-svn: 259574
* AMDGPU: Do not promote allocas with non-inbounds GEPsMatt Arsenault2016-02-026-76/+111
| | | | | | | | If we can't assume the pointer value isn't within the bounds of the object, it seems risky to try to replace the pointer calculations. llvm-svn: 259573
OpenPOWER on IntegriCloud