__unique_stable_name implementation. #16

isl_map_from_union_map cannot determine the map's space if the union_map is empty. polly::singleton was designed for this case. We pass the expected map space to avoid crashing in isl_map_from_union_map. This fixes an issue found by the aosp buildbot. Thanks to Eli Friedman for the reproducer. llvm-svn: 361290

llvm-svn: 361291

llvm-svn: 361292

Register coalescer fails for the test in the patch with the assertion in JoinVals::ConflictResolution `DefMI != nullptr'. It attempts to join live intervals for two adjacent instructions and erase the copy: %2:vreg_256 = COPY %1 %3:vreg_256 = COPY killed %1 The LI needs to be adjusted to kill subrange for the erased instruction and extend the subrange of the original def. That was done for the main interval only but not for the subrange. As a result subrange had a VNI pointing to the erased slot resulting in the above failure. Differential Revision: https://reviews.llvm.org/D62162 llvm-svn: 361293

Summary: Target link variables are currently implemented by creating a copy of the variables on the device side and unified memory never gets exploited. When the prgram uses the: ``` #pragma omp requires unified_shared_memory ``` directive in conjunction with a declare target link, the linked variable is no longer allocated on the device and the host version is used instead. This behavior is overridden by performing an explicit mapping. A Clang side patch is required. Reviewers: ABataev, AlexEichenberger, grokos, Hahnfeld Reviewed By: AlexEichenberger, grokos, Hahnfeld Subscribers: Hahnfeld, jfb, guansong, jdoerfert, openmp-commits Tags: #openmp Differential Revision: https://reviews.llvm.org/D60223 llvm-svn: 361294

Differential Revision: https://reviews.llvm.org/D62021 llvm-svn: 361295

Fixes PR41215 Differential Revision: https://reviews.llvm.org/D60283 llvm-svn: 361296

llvm-svn: 361297

…time Summary: This patch adds support for the registration of the requires directives with the runtime. Each requires directive clause will enable a particular flag to be set. The set of flags is passed to the runtime to be checked for compatibility with other such flags coming from other object files. The registration function is called whenever OpenMP is present even if a requires directive is not present. This helps detect cases in which requires directives are used inconsistently. Reviewers: ABataev, AlexEichenberger, caomhin Reviewed By: ABataev, AlexEichenberger Subscribers: jholewinski, guansong, jfb, jdoerfert, cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D60568 llvm-svn: 361298

…follow the same order as CSR saves in the prologue llvm-svn: 361299

This permits an init-capture to introduce a new pack: template<typename ...T> auto x = [...a = T()] { /* a is a pack */ }; To support this, the mechanism for allowing ParmVarDecls to be packs has been extended to support arbitrary local VarDecls. llvm-svn: 361300

CLANG_DEFAULT_STDLIB is defined to libstdc++. llvm-svn: 361301

llvm-svn: 361302

…ugh add/sub Looks we can transform all 8 variants of the pattern: https://rise4fun.com/Alive/auH This comes up as an issue on the path towards https://bugs.llvm.org/show_bug.cgi?id=41952 llvm-svn: 361303

For consistency with nearby code and to avoid interceptors during reports. llvm-svn: 361304

llvm-svn: 361305

llvm-svn: 361306

Enable CMake policy 77. This alters the behavior of option. The old behavior would remove the value of the option from the cache and create a new one. The new behavior does not create the variable if it is defined already. This ensures that subsequent reconfigures will behave identically. This seems better than the setting of OLD - the desire is to ensure that it is set to OLD or NEW. llvm-svn: 361307

llvm-svn: 361308

This change adds implementation to ompt_finalize_tool() and ompt_get_task_memory(). Patch by Hansang Bae Differential Revision: https://reviews.llvm.org/D61657 llvm-svn: 361309

Two tests having the same name creates a race condition when moving the trace files. llvm-svn: 361310

llvm-svn: 361311

Also, disallow specifying -no-pie/-nopie along with -static-pie. Differential Revision: https://reviews.llvm.org/D59841 llvm-svn: 361312

Double the number of files to merge, and use wait instead of sleep. llvm-svn: 361313

Summary: Values returned by GCCInstallation.getParentLibPath() and GCCInstallation.getTriple() are not valid unless GCCInstallation.isValid() returns true. This has previously been ignored, and the former two values were used without checking whether GCCInstallation is valid. This led to the bad path "/../bin" being added to the list of program paths. author: danielmentz "Daniel Mentz <[email protected]>" Reviewers: #clang, tstellar, srhines Reviewed By: srhines Subscribers: danielmentz, ormris, nickdesaulniers, srhines, cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D57930 llvm-svn: 361314

In these cases we would prefer a direct comparison over going through a vector type. llvm-svn: 361315

The SkipAppInitFiles setter was ignoring its import argument. llvm-svn: 361316

As discussed in D62024, we want to limit any potential IR transforms of shuffles to cases where we know the SDAG conversion would result in equivalent patterns for these IR variants. llvm-svn: 361317

llvm-svn: 361318

llvm-svn: 361319

llvm-svn: 361320

When I moved the resolve code from FileSpec to the FileSystem class, I introduced a regression. If you compare the two implementations, you'll notice that if the path doesn't exist, we should only reverse the effects of makeAbsolute, not the effects of tilde expansion. As a result, the logic to create the ~/.lldb directory broke, because we would resolve the path before creating it. Because the directory didn't exist yet, we'd call create_directories on the unresolved path, which failed. Differential revision: https://reviews.llvm.org/D62219 llvm-svn: 361321

…teJITDylib. JITDylibs should have unique names. This patch adds code to lli to respect this invariant (by refering to the exist JITDylib if a -jd <name> option is specified more than once). It also adds usage notes to the doxygen comment for createJITDylib method in ExecutionSession and LLJIT. http://llvm.org/PR41937 llvm-svn: 361322

llvm-svn: 361325

If process $PID has already exited, wait will give a non-zero exit status. llvm-svn: 361326

r360889 added new llround builtin functions. This patch adds their signatures for the WebAssembly backend. It also adds wasm32 support to utils/update_llc_test_checks.py, since that's the script other targets are using for their testcases for this feature. Differential Revision: https://reviews.llvm.org/D62207 llvm-svn: 361327

representing no such object, and an "Indeterminate" state representing an uninitialized object. The latter is not yet used, but soon will be. llvm-svn: 361328

of a union within constant expression evaluation. llvm-svn: 361329

llvm-svn: 361330

There should probably be nonconvergent versions, but my guess is it doesn't matter in practice. llvm-svn: 361331

llvm-svn: 361332

llvm-svn: 361333

Public and Private link libraries get merged in the LINK_LIBRARIES property instead of being kept separate. With any luck this will get `BUILD_SHARED_LIBS` working again on Linux. llvm-svn: 361334

Currently, a pragma AST node's recorded location starts at the namespace token (such as `omp` in the case of OpenMP) after the `#pragma` token, and the `#pragma` location isn't available. However, the `#pragma` location can be useful when, for example, rewriting a directive using Clang's Rewrite facility. This patch makes `#pragma` locations available in any `PragmaHandler` but it doesn't yet make use of them. This patch also uses the new `struct PragmaIntroducer` to simplify `Preprocessor::HandlePragmaDirective`. It doesn't do the same for `PPCallbacks::PragmaDirective` because that changes the API documented in `clang-tools-extra/docs/pp-trace.rst`, and I'm not sure about backward compatibility guarantees there. Reviewed By: ABataev, lebedev.ri, aaron.ballman Differential Revision: https://reviews.llvm.org/D61643 llvm-svn: 361335

Ensures that parent and all child processes are killed at once. llvm-svn: 361336

llvm-svn: 361337

This should be a valid exception to the general rule of not creating new shuffle masks in IR... because we already do it. :) Also, DAG combining/legalization will undo this by widening the shuffle back out if needed. Explanation for how we already do this: SLP or vector source can create chains of insert/extract as shown in 1 of the examples from PR16739: https://godbolt.org/z/NlK7rA https://bugs.llvm.org/show_bug.cgi?id=16739 And we expect instcombine or DAGCombine to clean that up by creating relatively simple shuffles. Differential Revision: https://reviews.llvm.org/D62024 llvm-svn: 361338

Ensure that log file has been fully updated before trying to read it. llvm-svn: 361339

These are needed to avoid undefined symbols which aren't satisfied by Clang itself. Differential Revision: https://reviews.llvm.org/D62174 llvm-svn: 361340

llvm-svn: 361341

CET-IBT enabled Return-twice functions will indirectly jump after the caller's position. So when CET-IBT is enable, we should make sure these is endbr* instructions follow these Return-twice function caller. Like GCC does. Patch by Xiang Zhang (xiangzhangllvm) Differential Revision: https://reviews.llvm.org/D61881 llvm-svn: 361342

Summary: When building Doxygen docs for llvm and clang, it helpfully prints a warning at the end noting that the `LOOKUP_CACHE_SIZE` value was too small to keep all symbols in memory. By increasing to the size it recommends, Doxygen builds have greatly improved performance. On my machine, time to run `doxygen-llvm` changes from 34 minutes to 22 minutes, which is a decent amount of time saved by changing a single number. Reviewed By: hintonda Patch by J. Ryan Stinnett! Tags: #clang, #llvm Differential Revision: https://reviews.llvm.org/D62138 llvm-svn: 361343

Summary: run-clang-tidy.py was enforcing '-header-filter' parameter with an unfortunate default value when none was given. Thus, leading to overwritten clang-tidy configuration (e.g. from .clang-tidy). This change removes the default value for '-header-filter' resulting in the default behaviour of clang-tidy itself. Fixes PR#41426 Reviewed By: hintonda Patch by Torbjörn Klatt! Tags: #clang-tools-extra, #clang Differential Revision: https://reviews.llvm.org/D61747 llvm-svn: 361344

llvm-svn: 361345

Differential Revision: https://reviews.llvm.org/D62173 llvm-svn: 361346

…rm - NFC llvm-svn: 361347

The MSVC CRT uses TLS storage to implement per-thread locales. This storage gets freed during program termination, and if we attempt to do any io operations (like flushing the std streams) after this occurs the program may abort. This patch is a speculative fix for that issue. The fix tries forcing the initialization of the locale TLS before initializing the std streams. This should mean that the TLS is freed after we destroy the streams. llvm-svn: 361348

llvm-svn: 361349

llvm-svn: 361350

Filters out noise, and distinguish Mach-O related options from others. Differential Revision: https://reviews.llvm.org/D62195 llvm-svn: 361351

…41971) Fix for https://bugs.llvm.org/show_bug.cgi?id=41971. Make the combineVectorSizedSetCCEquality() transform more conservative by checking that the bitcast to the vector type will be cheap/free for both operands. I'm considering it cheap if it's a constant, a load or already a vector. I've dropped the explicit check for f128 because it should fall out naturally (in the cases where it'd be detrimental). Differential Revision: https://reviews.llvm.org/D62220 llvm-svn: 361352

llvm-svn: 361353

The Armv8.2-A crypto extensions all defaulted to true, but should default to false, like all the other extensions. Differential Revision: https://reviews.llvm.org/D62180 llvm-svn: 361354

Summary: For big-endian powerpc64, the default ABI is ELFv1. OpenPower ABI ELFv2 is supported when -mabi=elfv2 is specified. FreeBSD support for PowerPC64 ELFv2 ABI with LLVM is in progress[1]. This patch adds an alternative way to specify ELFv2 ABI on target triple [2]. The following results are expected: ELFv1 when using: -target powerpc64-unknown-freebsd12.0 -target powerpc64-unknown-freebsd12.0 -mabi=elfv1 -target powerpc64-unknown-freebsd12.0-elfv1 ELFv2 when using: -target powerpc64-unknown-freebsd12.0 -mabi=elfv2 -target powerpc64-unknown-freebsd12.0-elfv2 [1] https://wiki.freebsd.org/powerpc/llvm-elfv2 [2] https://clang.llvm.org/docs/CrossCompilation.html Patch by Alfredo Dal'Ava Júnior! Differential Revision: https://reviews.llvm.org/D61950 llvm-svn: 361355

This is the second part of the commit fixing PR38917 (hoisting partitially redundant machine instruction). Most of PRE (partitial redundancy elimination) and CSE work is done on LLVM IR, but some of redundancy arises during DAG legalization. Machine CSE is not enough to deal with it. This simple PRE implementation works a little bit intricately: it passes before CSE, looking for partitial redundancy and transforming it to fully redundancy, anticipating that the next CSE step will eliminate this created redundancy. If CSE doesn't eliminate this, than created instruction will remain dead and eliminated later by Remove Dead Machine Instructions pass. The third part of the commit is supposed to refactor MachineCSE, to make it more clear and to merge MachinePRE with MachineCSE, so one need no rely on further Remove Dead pass to clear instrs not eliminated by CSE. First step: https://reviews.llvm.org/D54839 Fixes llvm.org/PR38917 llvm-svn: 361356

Summary: The wide-string equivalent of strdup. Implementation trivial. Reviewers: vitalybuka, eugenis Subscribers: kubamracek, delcypher, llvm-commits, #sanitizers Tags: #llvm, #sanitizers Differential Revision: https://reviews.llvm.org/D62189 llvm-svn: 361357

llvm-svn: 361358

…carded sections Rather than report "undefined symbol: ", give more informative message about the object file that defines the discarded section. In particular, PR41133, if the section is a discarded COMDAT, print the section group signature and the object file with the prevailing definition. This is useful to track down some ODR issues. We need to * add `uint32_t DiscardedSecIdx` to Undefined for this feature. * make ComdatGroups public and change its type to DenseMap<CachedHashStringRef, const InputFile *> Reviewed By: ruiu Differential Revision: https://reviews.llvm.org/D59649 llvm-svn: 361359

Summary: This patch introduces the DWARFTypeUnit class, and teaches lldb to parse type units out of both the debug_types section (DWARF v4), and from the regular debug_info section (DWARF v5). The most important piece of functionality - resolving DW_AT_signatures to connect type forward declarations to their definitions - is not implemented here, but even without that, a lot of functionality becomes available. I've added tests for the commands that start to work after this patch. The changes in this patch were greatly inspired by D61505, which in turn took over changes from D32167. Reviewers: JDevlieghere, clayborg, aprantl Subscribers: mgorny, jankratochvil, lldb-commits Differential Revision: https://reviews.llvm.org/D62008 llvm-svn: 361360

llvm-svn: 361361

Summary: Log the AST of the TU associated with LLDB's `expr` command, once a declaration is completed Reviewers: shafik Subscribers: rnkovacs, dkrupp, Szelethus, gamesh411, lldb-commits Tags: #lldb Differential Revision: https://reviews.llvm.org/D62061 llvm-svn: 361362

llvm-svn: 361363

Symbol's NameSize is computed lazily. Currently, when we replace a symbol, a cached length value can be discarded. This patch propagates that value. Differential Revision: https://reviews.llvm.org/D62234 llvm-svn: 361364

llvm-svn: 361365

llvm-svn: 361366

This adds proper handling of the NONAME-keyword, which makes llvm-dlltool generate an import using the ordinal instead of the name. Patch by by Jannik Vogel, test added by Stefan Schmidt. Differential Revision: https://reviews.llvm.org/D62175 llvm-svn: 361367

…nd -fcs-profile-generate Follows on from r360674 which added it for -fprofile-instr-generate. Differential Revision: https://reviews.llvm.org/D62200 llvm-svn: 361368

When the tiny code model is requested for a target machine that does not support this, we get an error message (which is nice) but also this diagnostic and request to submit a bug report: fatal error: error in backend: Target does not support the tiny CodeModel [Inferior 2 (process 31509) exited with code 0106] clang-9: error: clang frontend command failed with exit code 70 (use -v to see invocation) (gdb) clang version 9.0.0 (http://llvm.org/git/clang.git 29994b0c63a40f9c97c664170244a7bba5ecc15e) (http://llvm.org/git/llvm.git 95606fdf91c2d63a931e865f4b78b2e9828ddc74) Target: arm-arm-none-eabi Thread model: posix clang-9: note: diagnostic msg: PLEASE submit a bug report to https://bugs.llvm.org/ and include the crash backtrace, preprocessed source, and associated run script. clang-9: note: diagnostic msg: ******************** PLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT: Preprocessed source(s) and associated run script(s) are located at: clang-9: note: diagnostic msg: /tmp/tiny-dfe1a2.c clang-9: note: diagnostic msg: /tmp/tiny-dfe1a2.sh clang-9: note: diagnostic msg: But this is not a bug, this is a feature. :-) Not only is this not a bug, this is also pretty confusing. This patch causes just to print the fatal error and not the diagnostic: fatal error: error in backend: Target does not support the tiny CodeModel Differential Revision: https://reviews.llvm.org/D62236 llvm-svn: 361370

llvm-svn: 361371

…eration Overaligned and underaligned types (i.e. types where the alignment has been increased or decreased using the aligned and packed attributes) weren't being correctly handled in all cases, as the unadjusted alignment should be used. This patch also adjusts getTypeUnadjustedAlign to correctly handle typedefs of non-aggregate types, which it appears it never had to handle before. Differential Revision: https://reviews.llvm.org/D62152 llvm-svn: 361372

.debug_ranges parsing is not well tested [citation needed] because this section tends to be only used in optimized code, and we don't build optimized executables in our tests, as they produce unpredictable results. This patch aims to add some very simple tests for parsing static range data, which can serve as a first line of defense in case things break. I also include one XFAILed test, which demonstrates that we don't correctly handle mixed DWARF v5 and v4 ranges in a single file. llvm-svn: 361373

llvm-svn: 361374

Fixes scan-build warning. llvm-svn: 361375

Summary: Instead of failing with assertions. Fixes a crash found by oss-fuzz: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=12865 Reviewers: sammccall Reviewed By: sammccall Subscribers: cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D62137 llvm-svn: 361376

…tax library" Also reverted r361264 "[Syntax] Rename TokensTest to SyntaxTests. NFC" which built on it. This is because there were hitting an assert on bots http://lab.llvm.org:8011/builders/llvm-clang-lld-x86_64-scei-ps4-ubuntu-fast http://lab.llvm.org:8011/builders/llvm-clang-lld-x86_64-scei-ps4-windows10pro-fast llvm-svn: 361377

llvm-svn: 361378

… is not zero. NFCI. Fixes scan-build division by zero warning. llvm-svn: 361379

…re visible The check lines are being affected by an upcoming patch, regenerate the checklines to visualize the changes better. llvm-svn: 361380

Summary: Another target that prefers to use `-march` in tests ``` llvm/test/CodeGen/SPARC$ grep -ri mtriple | wc -l 25 llvm/test/CodeGen/SPARC$ grep -ri march | wc -l 165 ``` This test is being affected by a further patch, so regenerate it to better visualize the changes Reviewers: RKSimon, dcederman, gberry Reviewed By: RKSimon Subscribers: jyknight, fedor.sergeev, jrtc27, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D62242 llvm-svn: 361381

Support the OpenCL C pipe feature in C++ for OpenCL mode, to preserve backwards compatibility with OpenCL C. Various changes had to be made in Parse and Sema to enable pipe-specific diagnostics, so enable a SemaOpenCL test for C++. Differential Revision: https://reviews.llvm.org/D62181 llvm-svn: 361382

Summary: to test SVN commit access with a simple change. Reviewers: jankratochvil Subscribers: mgorny, lldb-commits Tags: #lldb Differential Revision: https://reviews.llvm.org/D62243 llvm-svn: 361383

This patch brings various error messages into line with each other, by removing trailing full stops, and making the first letter lower-case. This addresses https://bugs.llvm.org/show_bug.cgi?id=40859. Reviewed by: jhenderson, rupprecht, jakehehrlich Differential Revision: https://reviews.llvm.org/D62072 Patch by Alex Brachet llvm-svn: 361384

…file. Summary: This helps building internal tools on top of the library. Reviewers: gchatelet Subscribers: tschuett, llvm-commits, bdb, ondrasej Tags: #llvm Differential Revision: https://reviews.llvm.org/D62239 llvm-svn: 361385

See bug 41361: https://bugs.llvm.org/show_bug.cgi?id=41361 Reviewers: artem.tamazov, arsenm Differential Revision: https://reviews.llvm.org/D61012 llvm-svn: 361386

…non-zero loop cost. NFCI. The input LoopCost value can be zero, but if so it should be recalculated with the current VF. After that it should always be non-zero. llvm-svn: 361387

@umanwizard

Summary: Make it clear that the default is 100. Patch by Brennan Vincent(@umanwizard)! Reviewers: #clang-tools-extra, kadircet Reviewed By: kadircet Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, kadircet, cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D61865 llvm-svn: 361388

llvm-svn: 361389

…l permutation coverage Somehow missed some patterns initially.. While there, add comments. llvm-svn: 361390

Reverted in r361377. Also reland the '.gn' files (reverted in r361389). llvm-svn: 361391

…art enum. Transformer provides an enum to indicate the range of source text to be edited. That support is now redundant with the new (and more general) RangeSelector library, so we remove the custom enum support in favor of supporting any RangeSelector. Reviewers: ilya-biryukov Subscribers: cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D62149 llvm-svn: 361392

llvm-svn: 361393

llvm-svn: 361394

…ecified with --section This fixes PR41886: llvm-objdump -d -r -j .text doesn't show inline relocations of .text While here, switch to stable_sort() because we don't want to change the order of relocations applied to the same location. gABI says consecutive relocation records are composed together and their order matters. In practise it is difficult to see relocations applied to the same location not consecutive, we just have to keep the relative order of relocations with the same offset. Reviewed By: grimar Differential Revision: https://reviews.llvm.org/D62253 llvm-svn: 361395

Being affected by upcoming patch llvm-svn: 361396

Being affected by upcoming patch llvm-svn: 361397

… check-prefix Does not affect update_llc_test_checks, or the actual output, but is not accepted by the actual FileCheck. Sorry, i should have noticed this before committing, not the very next second after.. llvm-svn: 361398

llvm-svn: 361399

Subscribers: cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D62214 llvm-svn: 361400

This is a minimal start to correcting a problem most directly discussed in PR38086: https://bugs.llvm.org/show_bug.cgi?id=38086 We have been hacking around a limitation for FP select patterns by using the fast-math-flags on the condition of the select rather than the select itself. This patch just allows FMF to appear with the 'select' opcode. No changes are needed to "FPMathOperator" because it already includes select-of-FP because that definition is based on the (return) value type. Once we have this ability, we can start correcting and adding IR transforms to use the FMF on a 'select' instruction. The instcombine and vectorizer test diffs only show that the IRBuilder change is behaving as expected by applying an FMF guard value to 'select'. For reference: rL241901 - allowed FMF with fcmp rL255555 - allowed FMF with FP calls Differential Revision: https://reviews.llvm.org/D61917 llvm-svn: 361401

If a template parameter refers to a pointer to member, but the mangling of that was a string literal instead of a real symbol, llvm-undname used to crash instead of rejecting the input. llvm-svn: 361402

llvm-svn: 361403

…anType Summary: This extends Krzysztof Parzyszek's X86-specific solution (https://reviews.llvm.org/D60208) to the generic code pointed out by James Y Knight. Reviewers: kparzysz, craig.topper, nickdesaulniers Subscribers: efriedma, sdardis, nemanjai, javed.absar, eraman, fedor.sergeev, asb, rbar, johnrusso, simoncook, apazos, sabuasal, niosHD, jrtc27, zzheng, edward-jones, atanasyan, rogfer01, MartinMosbeck, brucehoult, the_o, PkmX, jocewei, jsji, llvm-commits, srhines, void, nickdesaulniers, jyknight Tags: #llvm Differential Revision: https://reviews.llvm.org/D60224 llvm-svn: 361404

Keep it optional in cases this is ever needed in some global context. Currently it's only used for getting an upper bound inline asm code size. For AMDGPU, gfx10 increases the maximum instruction size to 20-bytes. This avoids penalizing older subtargets when estimating code size, and making some annoying branch relaxation test adjustments. llvm-svn: 361405

Don't check for unsupported targets for every instruction. llvm-svn: 361406

Fuchsia has several exceptions which are merely informational and should not be treated as crashes. This patch puts logic that read from the exception port and handled the exception in a loop, and ignores non-exceptions and informational exceptions. Patch By: aarongreen Differential Revision: https://reviews.llvm.org/D62226 llvm-svn: 361407

…1584. Removed wrong debug assertion. Differential Revision: https://reviews.llvm.org/D62251 llvm-svn: 361408

…fold https://rise4fun.com/Alive/OaY llvm-svn: 361409

Summary: EH Frames aren't supported on AIX with the system compiler, but the definition of HAVE_EHTABLE_SUPPORT misses this which causes linking problems on AIX. This patch updates the definition of HAVE_EHTABLE_SUPPORT in both JITLink and RuntimeDyld. Author: daltenty Reviewers: sfertile, xingxue, hubert.reinterpretcase Reviewed By: xingxue Subscribers: hiraditya, jsji, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D62203 llvm-svn: 361410

llvm-svn: 361411

Rewrite the GetHistoryFilePath implementation without relying on FileSpec in the spirit of our discussion in D61994. It changes LLDBs behavior in two ways: 1. We now only use the -widehistory suffix when LLDB is built with wchar support, instead of as the fallback from when the ~/.lldb directory isn't writable. 2. When the ~/.lldb directory isn't writable, we don't write any history files at all. Previously we would write them to the user's home directory (with the incorrect wide suffix), polluting ~ with a different file for every IO handler. Differential revision: https://reviews.llvm.org/D62216 llvm-svn: 361412

Add support for creating a `StencilPart` from any `RangeSelector`, which broadens the scope of `Stencil`. Correspondingly, deprecate Stencil's specialized combinators `node` and `sNode` in favor of using the new `selection` combinator directly (with the appropriate range selector). Reviewers: sbenza Subscribers: cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D62160 llvm-svn: 361413

Some tests assume that iteration through an unordered multimap elements will return them in the same order as at the container creation. This assumption is not true since the container is unordered, so that no specific order of elements is ever guaranteed for such container. This patch is a continuation of D54838 and introduces checks verifying that any iteration will return elements exactly from a set of valid values and without repetition, but in no particular order. Thanks to Andrey Maksimov for the patch. Differential Revision: https://reviews.llvm.org/D56498 llvm-svn: 361414

llvm-svn: 361415

Summary: Avoid visiting an instruction more than once by using a map. Reviewers: davidxl Reviewed By: davidxl Subscribers: llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D62262 llvm-svn: 361416

This fix is for the problem from https://bugs.llvm.org/show_bug.cgi?id=38714. Specifically, Simple Register Coalescing creates following conversion : undef %0.sub_32:gpr64 = ORRWrs $wzr, %3:gpr32common, 0, debug-location !24; It copies 32-bit value from gpr32 into gpr64. But Live DEBUG_VALUE analysis is not able to create debug location record for that instruction. So the problem is in that debug info for argc variable is incorrect. The fix is to write custom isCopyInstrImpl() which would recognize the ORRWrs instr. llvm-svn: 361417

… rule. This revision introduces an adaptor from Transformer's rewrite rules (`clang::tooling::RewriteRule`) to `ClangTidyCheck`. For example, given a RewriteRule `MyCheckAsRewriteRule`, it lets one define a tidy check as follows: ``` class MyTidyCheck : public TransformerClangTidyCheck { public: MyTidyCheck(StringRef Name, ClangTidyContext *Context) : TransformerClangTidyCheck(MyCheckAsRewriteRule, Name, Context) {} }; ``` Reviewers: aaron.ballman Subscribers: mgorny, xazax.hun, cfe-commits, ilya-biryukov Tags: #clang Differential Revision: https://reviews.llvm.org/D61386 llvm-svn: 361418

llvm-svn: 361419

llvm-svn: 361420

Summary: Parallel level counter should be volatile to prevent some dangerous optimiations by the ptxas. Otherwise, ptxas optimizations lead to undefined behaviour in some cases. Also, use __threadfence() for #pragma omp flush and if the barrier should not be used (we have only one thread in the team), still perform flush operation since the standard requires implicit flush when executing barriers. Reviewers: gtbercea, kkwli0, grokos Subscribers: guansong, jfb, jdoerfert, openmp-commits, caomhin Tags: #openmp Differential Revision: https://reviews.llvm.org/D62199 llvm-svn: 361421

Using model numbers found in Table 2-1 of the May 2019 version of the Intel Software Developer's Manual Volume 4. llvm-svn: 361422

… __builtin_cpu_is. Using model numbers found in Table 2-1 of the May 2019 version of the Intel Software Developer's Manual Volume 4. llvm-svn: 361423

'thread_local' in C++20. llvm-svn: 361424

…nsics into llvm.ceil/floor. Remove some isel patterns that existed because that was happening. We were turning roundss/sd/ps/pd intrinsics with immediates of 1 or 2 into llvm.floor/ceil. The llvm.ceil/floor intrinsics are supposed to correspond to the libm functions. For the libm functions we need to disable the precision exception so the llvm.floor/ceil functions should always map to encodings 0x9 and 0xA. We had a mix of isel patterns where some used 0x9 and 0xA and others used 0x1 and 0x2. We need to be consistent and always use 0x9 and 0xA. Since we have no way in isel of knowing where the llvm.ceil/floor came from, we can't map X86 specific intrinsics with encodings 1 or 2 to it. We could map 0x9 and 0xA to llvm.ceil/floor instead, but I'd really like to see a use case and optimization advantage first. I've left the backend test cases to show the blend we now emit without the extra isel patterns. But I've removed the InstCombine tests completely. llvm-svn: 361425

The previous patch lost the call to PowerOf2Ceil, which causes LLD to crash when handling common symbols with a non-power-of-2 size. I tweaked the existing common.test to make the bsspad16 common symbol be 15 bytes to add coverage for this case. llvm-svn: 361426

Reviewers: dmajor, vitalybuka, waltl Subscribers: kubamracek, #sanitizers, llvm-commits Tags: #sanitizers, #llvm Differential Revision: https://reviews.llvm.org/D61871 llvm-svn: 361427

Found in a bootstrap of LLVM with implicit modules, resulting in a deadlock of some Orc unit tests with libstdc++ 8.1. An enum was used as part of the implementation of std::recursive_mutex and this bug resulted in the constant initialization of zero instead of the desired non-zero value. => Badness. Richard Smith tells me neither of these fields are necessarily canonical & so using declaresSamEntity is the right solution here (rather than changing both of these Fields to be canonical by construction/from their source) llvm-svn: 361428

Before e97b5f5 ([clang][Darwin] Refactor header search path logic into the driver), both --sysroot and -isysroot worked to specify where to look for system and C++ headers on Darwin. However, that change caused clang to start ignoring --sysroot. This fixes the regression, and adds tests. (I also note that on all other platforms, clang seems to almost completely ignore -isysroot, but that's another issue...) Differential Revision: https://reviews.llvm.org/D62268 llvm-svn: 361429

… time. http://lab.llvm.org:8011/builders/llvm-clang-x86_64-expensive-checks-win/builds/17792/steps/test-check-all/logs/stdio Failing Tests (1): LLVM :: CodeGen/AMDGPU/regbank-reassign.mir llvm-svn: 361430

…e of 0. Remove a bunch of isel patterns that become unnecessary. We effectively had a second set of isel patterns that tried to use a regular store instruction and an extract_subreg instruction. Or a masked move and an extract_subreg. These patterns were intended to override the matching of VEXTRACT instructions by taking advantage of the priority of the explicit immediate 0 for the index. This patch instaed just disables the immediate 0 matchin the VEXTRACT patterns. This each of the component pieces of the larger patterns will match by themselves. This found a bug of sorts were we didn't use 128-bit store for 512->128 extract on KNL. Its unclear what the right thing here should be. Using the vextract avoids constraining the register allocator to use xmm0-15. But it always results in a longer encoding if the register allocator ends up choosing xmm0-15 anyway. llvm-svn: 361431

…d include/c++ This change is a consequence of the discussion in "RFC: Place libs in Clang-dedicated directories", specifically the suggestion that libunwind, libc++abi and libc++ shouldn't be using Clang resource directory. Tools like clangd make this assumption, but this is currently not true for the LLVM_ENABLE_PER_TARGET_RUNTIME_DIR build. This change addresses that by moving the output of these libraries to lib/$target/c++ and include/c++ directories, leaving resource directory only for compiler-rt runtimes and Clang builtin headers. Differential Revision: https://reviews.llvm.org/D59168 llvm-svn: 361432

llvm-svn: 361433

This allows using anything that isn't a literal integer as the bounds for a foreach. Some of the diagnostics aren't perfect, but nobody ever accused tablegen of having good errors. For example, the existing wording suggests a bitrange is valid, but as far as I can tell this has never worked. Fixes bug 41958. llvm-svn: 361434

A very minimal demo of how to use the LLJIT class, along the lines of the old HowToUseJIT example. llvm-svn: 361435

Summary: If we are building the tests for the runtimes we should make them depend on gtest so that gtest is built and ready before we run any of the check-* targets. Reviewers: phosek, compnerd Reviewed By: compnerd Subscribers: mgorny, winksaville, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D62269 llvm-svn: 361436

llvm-svn: 361437

llvm-svn: 361438

llvm-svn: 361439

The patch in r359029 missed a few accessors and mutators. This patch also changes the lock to a recursive one as OptionValueFileSpecList::Clear() can be invoked from some of the other methods. llvm-svn: 361440

llvm-svn: 361441

Summary: From what I understand, it's possible for multiple threads to request a specific language runtime (e.g. CPPLanguageRuntime). This leads to a data race. Reviewers: jingham, JDevlieghere, compnerd, clayborg Differential Revision: https://reviews.llvm.org/D62169 llvm-svn: 361442

The Windows Code Generation model cannot generation code with the PIC relocation model - all code is implicitly position independent due to the DLL load slide that occurs if it is not loaded at the preferred base address. Invert the condition and inline the single use of the variable. This should also aid the WASM target. This significantly improves the state of the (swift) repl on Windows (and should aid in expression evaluation on Windows). llvm-svn: 361443

The driver passes by reference an uninitialized num_errors variable to RunCommandInterpreter. This should be fine, as it's supposed to be an output argument, but the reproducer instrumentation reads it in order to record the value of all the arguments to the function. This change fixes it by initializing num_errors to 0 before calling RunCommandInterpreter. llvm-svn: 361444

Summary: Terminology introduced by [[#]] blocks is confusing and does not integrate well with existing terminology. First, variables referred by [[]] blocks are called "pattern variables" while the text a CHECK directive needs to match is called a "CHECK pattern". This is inconsistent with variables in [[#]] blocks since [[#]] blocks are also found in CHECK pattern yet those variables are called "numeric variable". Second, the replacing of both [[]] and [[#]] blocks by the value of the variable or expression they contain is represented by a FileCheckPatternSubstitution class. The naming refers to being a substitution in a CHECK pattern but could be wrongly understood as being a substitution of a pattern variable. Third and lastly, comments use "numeric expression" to refer both to the [[#]] blocks as well as to the numeric expressions these blocks contain which get evaluated at match time. This patch solves these confusions by - calling variables in [[]] and [[#]] blocks as string and numeric variables respectively; - referring to [[]] and [[#]] as substitution *blocks*, with the former being a string substitution block and the latter a numeric substitution block; - calling [[]] and [[#]] blocks to be replaced by the value of a variable or expression they contain a substitution (as opposed to definition when these blocks are used to defined a variable), with the former being a string substitution and the latter a numeric substitution; - renaming the FileCheckPatternSubstitution as a FileCheckSubstitution class with FileCheckStringSubstitution and FileCheckNumericSubstitution subclasses; - restricting the use of "numeric expression" to refer to the expression that is evaluated in a numeric substitution. While numeric substitution blocks only support numeric substitutions of numeric expressions at the moment there are plans to augment numeric substitution blocks to support numeric definitions as well as both a numeric definition and numeric substitution in the same numeric substitution block. Reviewers: jhenderson, jdenny, probinson, arichardson Subscribers: hiraditya, arichardson, probinson, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D62146 llvm-svn: 361445

Summary: With now a clear distinction between string and numeric substitutions, this patch introduces separate classes to represent them with a parent class implementing the common interface. Diagnostics in printSubstitutions() are also adapted to not require knowing which substitution is being looked at since it does not hinder clarity and makes the implementation simpler. Reviewers: jhenderson, jdenny, probinson, arichardson Subscribers: llvm-commits, probinson, arichardson, hiraditya Tags: #llvm Differential Revision: https://reviews.llvm.org/D62241 llvm-svn: 361446

LLDB -> liblldbcore.a llvm-svn: 361447

…rk mode if -collect_data_flow= is given llvm-svn: 361448

llvm-svn: 361449

Summary: This patch adds the file names to llvm-objcopy error messages. It makes easy to identify which file causes an error. Bugzilla: https://bugs.llvm.org/show_bug.cgi?id=41798 Reviewers: espindola, alexshap, rupprecht, jhenderson, jakehehrlich Reviewed By: rupprecht, jhenderson, jakehehrlich Subscribers: emaste, arichardson, jakehehrlich, MaskRay, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D61993 llvm-svn: 361450

…VE extensions This patch updates assembler attributes for AArch64 targets so we can disassemble newer instructions supported in ISA version 8.5 and SVE extensions. Differential Revision: https://reviews.llvm.org/D62235 llvm-svn: 361451

…nctionality is now part of libFuzzer proper; also write functions.txt to the disk only if this file doesn't exist yet llvm-svn: 361452

In general dynamic/local dynamic TLS models, with -fno-plt, * x86: emit `calll *___tls_get_addr@GOT(%ebx)` instead of `calll ___tls_get_addr@PLT` Note, on x86, if we can get rid of %ebx as the PIC register, it may be better to use a register not preserved across function calls. * x86_64: emit `callq *__tls_get_addr@GOTPCREL(%rip)` instead of `callq __tls_get_addr@PLT` Reorganize the code by separating 32-bit and 64-bit. Reviewed By: rnk Differential Revision: https://reviews.llvm.org/D62106 llvm-svn: 361453

Summary: In this patch, `ISD::RETURNADDR` is lowered on the emscripten target to the new Emscripten runtime function `emscripten_return_address`, which implements the functionality. Patch by Guanzhong Chen Reviewers: tlively, aheejin Reviewed By: tlively Subscribers: dschuff, sbc100, jgravelle-google, hiraditya, sunfish, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D62210 llvm-svn: 361454

llvm-svn: 361455

This ensures that whether the user uses short or cannonical version of the triple, Clang will still find the runtimes under the cannonical triple name. Differential Revision: https://reviews.llvm.org/D52132 llvm-svn: 361456

It's uncommon to rely on temporary lifetime extension when having a regular, non-`const&` value behaves identically. Since `Twine::str` and `buildFixMsgForStringFlag` both return regular `std::string`s, there's seemingly no point in having `const&` here. llvm-svn: 361457

Replaces the remaining C-style casts with explicit casts in Utility. The motivation is that they are (1) easier to spot and (2) don't have multiple meanings. llvm-svn: 361458

Fix two functions where we were passing FileSpecs by value, while we could pass by const reference. llvm-svn: 361459

Summary: Allow struct fields SRA and dead stores. This works by considering fields accesses from getElementPtr to be considered as a possible pointer root that can be cleaned up. We check that the variable can be SRA by recursively checking the sub expressions with the new isSafeSubSROAGEP function. basically this allows the array in following C code to be optimized out struct Expr { int a[2]; int b; }; static struct Expr e; int foo (int i) { e.b = 2; e.a[i] = 1; return e.b; } Reviewers: greened, bkramer, nicholas, jmolloy Reviewed By: jmolloy Subscribers: llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D61911 llvm-svn: 361460

Summary: Uses a heuristic to detect std::function and friends. Reviewers: kadircet Reviewed By: kadircet Subscribers: cfe-commits Tags: #clang Differential Revision: https://reviews.llvm.org/D62238 llvm-svn: 361461

The previous patch added a member set to store instructions that we could allow to wrap. But this wasn't cleared between searches meaning that they could get promoted, incorrectly, during the promotion of a separate valid chain. Differential Revision: https://reviews.llvm.org/D62254 llvm-svn: 361462

@clayborg

In D61502#1503247 @clayborg suggested that DWARFUnit *+dw_offset_t can be now replaced by DWARFDIE. It is moved from DWARFDebugInfoEntry to DWARFDIE as noted by @clayborg. I have also removed return type as (1) it was wrong in one case and (2) no existing caller used the return type. I also refactored the deep nesting noted by @JDevlieghere. Differential Revision: https://reviews.llvm.org/D62211 llvm-svn: 361463

Result of running 'npm audit fix', which tracks security vulnerabilities. llvm-svn: 361464

In details: 1. Content in clang offload bundler changed type to Expected<StringRef> 2. Unnecessary error check for unqualified-id was removed from SemaSYCL/address-space-parameter-conversions.cpp respecting the appropriate change from the community. Signed-off-by: Dmitry Sidorov <[email protected]>

Signed-off-by: Dmitry Sidorov <[email protected]>

This patch fixes 2 constructors, fixes get_count() and get_size() methods, adds operator== and operator!=, fixes hash implementation, adds static_asssert to range and id classes to have early check for wrong dimension, fixes few other errors. Signed-off-by: Vyacheslav N Klochkov <[email protected]>

Clang doesn't allow casting address spaces using reinterpret_cast. Replaced with C-style cast. Signed-off-by: Vlad Romanov <[email protected]>

Wait for event in get_profiling_info. Signed-off-by: Vlad Romanov <[email protected]>

Updated event::get_wait_list, event::wait_and_throw and event::wait to use new scheduler. Signed-off-by: Vlad Romanov <[email protected]>

Signed-off-by: Alexey Bader <[email protected]>

Clean-up after 8a95712. [SYCL] Get sampler and accessor types without using named fields of kernel object Using the types of the parameter variables of __init method instead Signed-off-by: Alexey Bader <[email protected]>

Signed-off-by: Alexey Sachkov <[email protected]>

Since this attribute can also be used in SYCL, it is better to remove 'OpenCL' prefix. Also added LangOpt's to the attribute Signed-off-by: Alexey Sachkov <[email protected]>

This fixes an issue with hanging on several OpenCL devices in the cases when we firstly enqueue command which is blocked, then enqueue another command which is not and wait for the second. More details in code comment. Signed-off-by: Vlad Romanov <[email protected]>

Some tests may hang which can cause failing of overall check-all target without clear message. Timeout will be set only if psutil python module exists, no timeout otherwise. Signed-off-by: Vlad Romanov <[email protected]>

Get profiling info is not supported on host device. Signed-off-by: Vlad Romanov <[email protected]>

Signed-off-by: Mariya Podchishchaeva <[email protected]>

event_t and sampler_t type names are replaced with __ocl_event_t and __ocl_sampler_t to avoid potential collisions with user types. This change allows to re-use OpenCL event and sampler types for SYCL and replace OpTypeEvent and OpTypeSampler with __ocl_event_t and __ocl_sampler_t in SYCL headers. OpenCL diagnostics is not enabled. Signed-off-by: Mariya Podchishchaeva <[email protected]>

This change allows to avoid re-flower pass using. Signed-off-by: Mariya Podchishchaeva <[email protected]>

Replace SPIR-V type names with built-in OpenCL type names. Use mangled class names in IR checking. Signed-off-by: Mariya Podchishchaeva <[email protected]>

Signed-off-by: Vladimir Lazarev <[email protected]>

…ct, /EHsc Signed-off-by: Vyacheslav N Klochkov <[email protected]>

Signed-off-by: Dmitry Sidorov <[email protected]>

…load, and removed the overload. This is no longer present in the SYCL 1.2.1 specification since revision 5. Signed-off-by: Paul Keir <[email protected]>

Signed-off-by: Jeff Hammond <[email protected]>

Signed-off-by: Vyacheslav N Klochkov <[email protected]>

getOffloadingDeviceToolChain is the equivalent of getToolChain for device ToolChains. It may be possible to merge the two but as Offloading Devices require some extra information (like host/device triple and OffloadKind) it may get a little messy. The switch statement in getToolChain also already seems a little complex. So it may not be worth merging and over complicating them. In general using getOffloadingDeviceToolChain moves some of the shared logic for device ToolChain selection for SYCL/CUDA/HIP/OpenMP into the function call, so it refactors it a bit. But in a more specific application to the SYCL implementation this call allows a ToolChain implementer to add their own ToolChain in place of the SYCL ToolChain when targeting a specific host / device triple. However, if they did not want to add their own ToolChain and simply wanted to add a new tool to the existing SYCL ToolChain they'd still be able to do that via overloading/adding to the GetTools method inside of the SYCLToolChain. So the intent is just to add more flexibility, rather than replace some functionality with some other functionality. Signed-off-by: Andrew Gozillon <[email protected]>

…YCL target No longer will the switch statement assert with llvm_unreachable, instead the appropriate Clang diagnostic warning for an invalid SYCL output should be emitted. This occurs in the initialize() call for the SYCL offloader/phase and action generator, similar to how CUDA emits errors for incorrect GPU Arch targets. Doing it similarly to OpenMP (inside CompilerInvocation.cpp) doesn't work in this case as if we pass an invalid target we won't accidentally generate an invalid/incorrect ToolChain like GetToolChain's we will instead return a nullptr which will ICE the compiler before it reaches CompilerInvocation.cpp, so the warning interaction has to be slightly different in this case. Also removing some unused variables I missed in the first commit. Signed-off-by: Andrew Gozillon <[email protected]>

This should be an NFC. Moved the diagnostic from the OffloadingActionBuilder's initialize method to the Driver's CreateOffloadingDeviceToolChains call. No need to track it in two places, invalid target caught as early as required. Should be no chance of ICE'ng the compiler by accessing a non-existant ToolChain. Signed-off-by: Andrew Gozillon <[email protected]>

New test case aims to showcase that existing architecture triples that are not valid for SYCL offloading also get rejected. In this case, just testing using x86_64. Signed-off-by: Andrew Gozillon <[email protected]>

Made the names returned from the get_info method and from the integration header of the same kernel equal. clGetKernelInfo(used by get_info) function returns size of kernel name + 1 and kernel name with additional symbol '\0'. So size of the string returned from get_info is one character longer and clGetKernelInfo writes the additional symbol into raw pointer of the string. But inegration header returns kernel name aka string("kernelname"). Respectively: """ std::string str(/*strlen("kernelname")*/10 + 1, ' '); strcpy((char*)str.c_str(), "kernelname"); str != std::string("kernelname"); """ Signed-off-by: Alexey Voronov <[email protected]>

Signed-off-by: Alexey Voronov <[email protected]>

Removed set_arg method calls from accessor constructors. Fixed require method: now will make dependencies in the scheduler. Added support set_arg for non source SYCL kernels. Signed-off-by: Alexey Voronov <[email protected]>

Signed-off-by: Alexey Voronov <[email protected]>

Merge remote-tracking branch 'intel' into 'sycl'

llvm.loop.[ii/max_concurrency] -> llvm.loop.[ii/max_concurrency].count Signed-off-by: Dmitry Sidorov <[email protected]>

Allow for compilation using multiple objects and also enable compilation from source. Initial support for compilation from fat archives only allowed the compile to be from a single object and the fat archive. These changes expand on that to allow compilation direct from source files and multiple objects. Signed-off-by: Michael D Toguchi <[email protected]>

Fixed incorrect storing and extractiong of global/constant accessor in set_arg(s) logic. Signed-off-by: Alexey Voronov <[email protected]>

Signed-off-by: Vyacheslav N Klochkov <[email protected]>

This code is redundant since we abandoned re-flower pass and OclCxx classes. Signed-off-by: Mariya Podchishchaeva <[email protected]>

The main problem that half type is defined as _Float16 on the device and as manually implemented type on host so compiler could add _Float16 to the integration header and it will produce errors in host compilation. Signed-off-by: Mariya Podchishchaeva <[email protected]>

Signed-off-by: Vyacheslav N Klochkov <[email protected]>

This change is preparing vec and SwizzleOp classes to replacement of getNumElements() constexpr method with sizeof...(Indices) and replacement of constexpr Indexer() method with template struct Indexer to workaround an error in MSVC compiler on Windows. Signed-off-by: Vyacheslav N Klochkov <[email protected]>

Rename SYCL specific loop attributes to avoid potential name conflicts. Signed-off-by: Dmitry Sidorov <[email protected]>

The fix adds a workaround for MSVC bug that is fixed only in v19.20, and enables vec and SwizzleOp classes compilation for Windows. The versions of MSVC older than v19.20 do not recognize constexpr functions getNumElements() and Indexer() as const and thus do not let using them in template class parameters. The fix inlines getNumElements() and turns constexpr function Indexer(int I) into template struct Indexer<int I>. Signed-off-by: Vyacheslav N Klochkov <[email protected]>

Signed-off-by: Vyacheslav N Klochkov <[email protected]>

Signed-off-by: Mariya Podchishchaeva <[email protected]>

Signed-off-by: Alexey Sachkov <[email protected]>

Signed-off-by: Mariya Podchishchaeva <[email protected]>

Signed-off-by: Artur Gainullin <[email protected]>

…build Signed-off-by: Vyacheslav N Klochkov <[email protected]>

Signed-off-by: Alexey Voronov <[email protected]>

Removed comments are redundant and complicate maintenance. Signed-off-by: Alexey Bader <[email protected]>

Signed-off-by: Alexey Bader <[email protected]>

Selectively enabled a few OpenCL diagnostics for sampler type as some diagnostics trigger on SYCL use cases. For instance, OpenCL kernel in SYCL mode initializes lambda captures with the kernel argument values. This initialization code for sampler emits errors because OpenCL disallows sampler on the right hand side of the binary operators - these are supposed to be used only by built-in functions. Another potential issue is the lambda object itself - captured sampler is a member of the lambda object and OpenCL doesn't allow composite types with samplers. SPIR-V produced from SYCL should be okay as lambda object can be removed by standard LLVM transformation passes. Signed-off-by: Alexey Bader <[email protected]>

to filter files with prefix match from dependency output. In sycl compilation, clang invokes cc1 two times, once for the host compilation and once for the offload compilation. The offload compilation generates a .h file, the host compilation includes this generated .h file. Finally, the clang driver deletes the generated temporary .h file before terminating. This new cc1 option allows the clang driver to specify that the temporary .h files shouldn't be added into the dependency information. Signed-off-by: Melanie Blower <[email protected]>

Signed-off-by: Artur Gainullin <[email protected]>

Signed-off-by: Vlad Romanov <[email protected]>

Rename global_mem_cache_type::write_only to global_mem_cache_type::read_write. Signed-off-by: Vlad Romanov <[email protected]>

Stream class wraps an accessor to the underlying stream buffer. Offset in this buffer is updated atomically to allow streaming by multiple threads safely. Current implementation is not complete. It supports output of string, char and boolean types. Other operators are defined with todo in the body. Signed-off-by: Artur Gainullin <[email protected]>

Community pull down introduced Expected<T> on the return of getContents. Apply the needed check after return to satisfy. From the llvm manual: All Error instances, whether success or failure, must be either checked or moved from (via std::move or a return) before they are destructed Signed-off-by: Michael D Toguchi <[email protected]>

* Warnings were fixed. * Bug in assertion was fixed. Signed-off-by: Artur Gainullin <[email protected]>

Signed-off-by: Vladimir Lazarev <[email protected]>

There was strict-aliasing rules violation. It could produce undefined behaviour. Signed-off-by: Mariya Podchishchaeva <[email protected]>

Signed-off-by: Jeff Hammond <[email protected]>

Also added check that passed SYCL context represents the same underlying OpenCL context associated with the OpenCL kernel object. Signed-off-by: Mariya Podchishchaeva <[email protected]>

Generate a wrapped device binary that can be used independently with a host link step. Takes the provided fat objects or source, extracts or creates the associated device binaries and wraps them to be used during a regular host link step Signed-off-by: Michael D Toguchi <[email protected]>

Use 0-dim accessors for scalar reductions . For buffers, reduction variables inherit dimensionality from accessors. For USM, reduction variables are assumed scalar without use of span. Signed-off-by: John Pennycook <[email protected]>

Even when compiling for the device, code meant only for host execution must be semantically analyzable. Existing wrapper headers were initially created to resolve this problem. But now -aux-triple is used to pass target triple of opposite side of compilation. Clang knows target triple for both sides of compilation so that preprocessor macros and target builtins from both sides are available. That is why there is no need in existing wrapper headers with stub definitions since they break compilation for cases when standard headers are included. Approach with -aux-triple is not complete solution. Host headers might provide definitions relying on host inline asm, intrinsics, using types not lowerable in device code. So we will need to provide a solution to overload some definitions for the device or to define only some macro for host headers. This will be resolved in other commits. Signed-off-by: Artur Gainullin <[email protected]>

Remove OpenCL cl_mem_fence_flags type/values definitions as the device compiler uses SPIR-V type/values now instead of OpenCL ones. Signed-off-by: Vyacheslav N Klochkov <[email protected]>

This change unifies handling of sampler and accessor types during OpenCL kernel generation. Signed-off-by: Mariya Podchishchaeva <[email protected]>

Signed-off-by: Vlad Romanov <[email protected]>

1) Fix error: "'detail::getSyclObjImpl': not a function". 2) Fix error: "subscript requires array or pointer type" for std::array use. 3) Fix few warnings caused by inconsistent usage of class-vs-struct keyword. Signed-off-by: Vyacheslav N Klochkov <[email protected]>

Patch 1: Implementation of Constructors for image class. Files Added: image.hpp - Includes the interface for image class. detail/image_impl.hpp - Includes the implementation of image_impl class. detail/image_impl.cpp - For now includes the utility functions. Will be expanded later as accessor and other APIs are implemented. test/basic_tests/image_constructors.cpp - test to check the image constructors. Next To Do: SPIRV Code generation and accessors implementations. Signed-off-by: Garima Gupta <[email protected]>

Signed-off-by: Kurt Chen <[email protected]> Signed-off-by: Vladimir Lazarev <[email protected]>

- Add development process description to contributing guide - Add info about skipping LIT tests targetting OpenCL device if corresponding RT is missed. Signed-off-by: Vladimir Lazarev <[email protected]>

According to C++ spec: Each enumerator becomes a named constant of the enumeration's type (that is, name), visible in the enclosing scope, and can be used whenever constants are required. Having unscoped enumerations defined in global namespace reserve some names and prohibits it's using in user's applications. For example, 'Device' might be used as user's class name and it might cause ambiguity. This patch aligns declaration of enums in spirv_types.hpp with ones from SPIRV-Headers project to avoid declaring identifiers in global namespace and simplify transition in the future. There are few exceptions: * uint32_t is used as base type for enum instead of unsigned. * '__spv' namespace is used instead of 'spv'. Signed-off-by: Alexey Sachkov <[email protected]>

Signed-off-by: Alexey Sachkov <[email protected]>

Remove CMake configuration expecting llvm-spirv and SYCL projects to be located in LLVM source tree. Location of external projects should be set via CMake variables. Signed-off-by: Alexey Bader <[email protected]>

Current API spelling is get_linear, which is wrong - should be get_linear_id according to spec. Fix the unit test as well. Signed-off-by: Konstantin Bobrovsky [email protected]

This implementation uses the keyword __unique_stable_name as an operator that can take a type or expression, and will result in a constexpr constant character array that (hopefully) uniquely represents the type or type-of-expression being passed to it. The unique representation is the normal mangled name of the type, except in the cases of a lambda, where the ordinal "number" is just replaced with LINE->COL. Macro expansions are also appended with '~' followed by the LINE->COL that it appears in. For example, a lambda declared in 'main' would look like _ZTSZ4mainEUlvE25->12 (line 25, column 12). A Lambda that comes from a macro looks like: _ZTSZ4mainEUlvE45->3~18->32 (macro invoked on 45/3, lambda defined inside the macro on line 18, column 32). Template instantiations based on the lambda would result in a name that contains the lambda mangling, for example: _ZTSZ3bazIZ4mainEUlvE25->12EvvEUlvE14->12 A function template named 'baz' that is instantiated with a lambda declared in 'main' on line 25/col 12 has another macro in it, declared on line 14, column 12. WIP Signed-off-by: Erich Keane <[email protected]>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

__unique_stable_name implementation. #16

__unique_stable_name implementation. #16

Commits on May 24, 2019

Commits on May 26, 2019

Commits on May 27, 2019

Commits on May 28, 2019

Commits on May 29, 2019

Commits on May 30, 2019

Commits on May 31, 2019

Commits on Jun 1, 2019

Commits on Jun 2, 2019

Commits on Jun 3, 2019

Commits on Jun 7, 2019

Commits on Jun 8, 2019

Commits on Jun 9, 2019

Commits on Jun 10, 2019

Commits on Jun 12, 2019

Commits on Jun 13, 2019

Commits on Jun 14, 2019

Commits on Jun 15, 2019

Commits on Jun 18, 2019

Commits on Jun 19, 2019

Commits on Jun 20, 2019

Commits on Jun 21, 2019

Commits on Jun 24, 2019