Kernel Fusion support for CUDA backend #2

sommerlukas · 2023-02-15T13:26:28Z

Extend kernel fusion for the CUDA backend.

In contrast to the existing SPIR-V based backends, the default binary format for the CUDA backend (PTX or CUBIN) is not suitable as input for the kernel fusion JIT compiler.

This PR therefore extends the driver to additionally embed LLVM IR in the fat binary if the user specifies the -fsycl-embed-ir during compilation, by taking the output of the sycl-post-link step for the CUDA backend.

The JIT compiler has been extended to handle LLVM IR as input format and PTX assembly as output format (including translation via the NVPTX backend). Target-specific parts of the fusion process have been refactored to TargetFusionInformation.

The connecting logic to the JIT compiler in the SYCL RT has been extended to produce valid PI device binaries for the CUDA backend/PI.

Heterogeneous ND ranges are not yet supported for the CUDA backend.

This PR fixes lit failures when building with `clang`. Since intel@9ac759f, when building with `clang`, there are two lit failures: ``` Failed Tests (2): cgeist :: Verification/memrefaddassign.cpp cgeist :: Verification/ptrarith.c ``` They crash with traceback: ``` PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace. Stack dump: 0. Program arguments: /localdisk2/waihungt/llvm/build/bin/cgeist /localdisk2/waihungt/llvm/polygeist/tools/cgeist/Test/Verification/ptrarith.c --function=* -S 1. <eof> parser at end of file #0 0x00007f1836120043 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/localdisk2/waihungt/llvm/build/bin/../lib/libLLVMSupport.so.16git+0x1e8043) #1 0x00007f183611de5e llvm::sys::RunSignalHandlers() (/localdisk2/waihungt/llvm/build/bin/../lib/libLLVMSupport.so.16git+0x1e5e5e) #2 0x00007f183612052f SignalHandler(int) Signals.cpp:0:0 #3 0x00007f1849701b20 __restore_rt sigaction.c:0:0 intel#4 0x00000000004d9a6e ValueCategory::SubIndex(mlir::OpBuilder&, mlir::Location, mlir::Type, mlir::Value, bool) const (/localdisk2/waihungt/llvm/build/bin/cgeist+0x4d9a6e) intel#5 0x00000000004da11f ValueCategory::GEPOrSubIndex(mlir::OpBuilder&, mlir::Location, mlir::Type, mlir::ValueRange, bool) const (/localdisk2/waihungt/llvm/build/bin/cgeist+0x4da11f) intel#6 0x00000000004da1bd ValueCategory::InBoundsGEPOrSubIndex(mlir::OpBuilder&, mlir::Location, mlir::Type, mlir::ValueRange) const (/localdisk2/waihungt/llvm/build/bin/cgeist+0x4da1bd) intel#7 0x00000000004b0394 MLIRScanner::EmitCheckedInBoundsPtrOffsetOp(mlir::Type, ValueCategory, mlir::ValueRange, bool, bool) (/localdisk2/waihungt/llvm/build/bin/cgeist+0x4b0394) ``` Signed-off-by: Tsang, Whitney <whitney.tsang@intel.com>

clang/include/clang/Driver/Action.h

clang/lib/Driver/ToolChains/Clang.cpp

sycl-fusion/jit-compiler/include/Options.h

sycl-fusion/jit-compiler/include/JITContext.h

sycl-fusion/passes/target/TargetFusionInfo.h

sycl/source/detail/jit_compiler.cpp

sycl-fusion/passes/target/TargetFusionInfo.cpp

sycl-fusion/jit-compiler/lib/translation/KernelTranslation.cpp

sycl/source/detail/scheduler/commands.hpp

Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

Avoid the dependencies of the dependencies to be removed when cleaning up the input commands from the graph without executing them. Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

sommerlukas · 2023-03-23T08:43:14Z

Closing this PR, as it is superseded by intel#8747.

This change prevents rare deadlocks observed for specific macOS/iOS GUI applications which issue many `dlopen()` calls from multiple different threads at startup and where TSan finds and reports a race during startup. Providing a reliable test for this has been deemed infeasible. Although I've only observed this deadlock on Apple platforms, conceptually the cause is not confined to Apple code so the fix lives in platform-independent code. Deadlock scenario: ``` Thread 2 | Thread 4 ReportRace() | Lock internal TSan mutexes | &ctx->slot_mtx | | dlopen() interceptor | OnLibraryLoaded() | MemoryMappingLayout::DumpListOfModules() | calls dyld API, which takes internal lock | lock() interceptor | TSan tries to take internal mutexes again | &ctx->slot_mtx call into symbolizer | MemoryMappingLayout::DumpListOfModules() calls dyld API, which hangs on trying to take lock ``` Resulting in: * Thread 2 has internal TSan mutex, blocked on dyld lock * Thread 4 has dyld lock, blocked on internal TSan mutex The fix prevents this situation by not intercepting any of the calls originating from `MemoryMappingLayout::DumpListOfModules()`. Stack traces for deadlock between ReportRace() and dlopen() interceptor: ``` thread #2, queue = 'com.apple.root.default-qos' frame #0: libsystem_kernel.dylib frame #1: libclang_rt.tsan_osx_dynamic.dylib`::wrap_os_unfair_lock_lock_with_options(lock=<unavailable>, options=<unavailable>) at tsan_interceptors_mac.cpp:306:3 frame #2: dyld`dyld4::RuntimeLocks::withLoadersReadLock(this=0x000000016f21b1e0, work=0x00000001814523c0) block_pointer) at DyldRuntimeState.cpp:227:28 frame #3: dyld`dyld4::APIs::_dyld_get_image_header(this=0x0000000101012a20, imageIndex=614) at DyldAPIs.cpp:240:11 frame intel#4: libclang_rt.tsan_osx_dynamic.dylib`__sanitizer::MemoryMappingLayout::CurrentImageHeader(this=<unavailable>) at sanitizer_procmaps_mac.cpp:391:35 frame intel#5: libclang_rt.tsan_osx_dynamic.dylib`__sanitizer::MemoryMappingLayout::Next(this=0x000000016f2a2800, segment=0x000000016f2a2738) at sanitizer_procmaps_mac.cpp:397:51 frame intel#6: libclang_rt.tsan_osx_dynamic.dylib`__sanitizer::MemoryMappingLayout::DumpListOfModules(this=0x000000016f2a2800, modules=0x00000001011000a0) at sanitizer_procmaps_mac.cpp:460:10 frame intel#7: libclang_rt.tsan_osx_dynamic.dylib`__sanitizer::ListOfModules::init(this=0x00000001011000a0) at sanitizer_mac.cpp:610:18 frame intel#8: libclang_rt.tsan_osx_dynamic.dylib`__sanitizer::Symbolizer::FindModuleForAddress(unsigned long) [inlined] __sanitizer::Symbolizer::RefreshModules(this=0x0000000101100078) at sanitizer_symbolizer_libcdep.cpp:185:12 frame intel#9: libclang_rt.tsan_osx_dynamic.dylib`__sanitizer::Symbolizer::FindModuleForAddress(this=0x0000000101100078, address=6465454512) at sanitizer_symbolizer_libcdep.cpp:204:5 frame intel#10: libclang_rt.tsan_osx_dynamic.dylib`__sanitizer::Symbolizer::SymbolizePC(this=0x0000000101100078, addr=6465454512) at sanitizer_symbolizer_libcdep.cpp:88:15 frame intel#11: libclang_rt.tsan_osx_dynamic.dylib`__tsan::SymbolizeCode(addr=6465454512) at tsan_symbolize.cpp:106:35 frame intel#12: libclang_rt.tsan_osx_dynamic.dylib`__tsan::SymbolizeStack(trace=StackTrace @ 0x0000600002d66d00) at tsan_rtl_report.cpp:112:28 frame intel#13: libclang_rt.tsan_osx_dynamic.dylib`__tsan::ScopedReportBase::AddMemoryAccess(this=0x000000016f2a2a90, addr=4381057136, external_tag=<unavailable>, s=<unavailable>, tid=<unavailable>, stack=<unavailable>, mset=0x00000001012fc310) at tsan_rtl_report.cpp:190:16 frame intel#14: libclang_rt.tsan_osx_dynamic.dylib`__tsan::ReportRace(thr=0x00000001012fc000, shadow_mem=0x000008020a4340e0, cur=<unavailable>, old=<unavailable>, typ0=1) at tsan_rtl_report.cpp:795:9 frame intel#15: libclang_rt.tsan_osx_dynamic.dylib`__tsan::DoReportRace(thr=0x00000001012fc000, shadow_mem=0x000008020a4340e0, cur=Shadow @ x22, old=Shadow @ 0x0000600002d6b4f0, typ=1) at tsan_rtl_access.cpp:166:3 frame intel#16: libclang_rt.tsan_osx_dynamic.dylib`::__tsan_read8(void *) at tsan_rtl_access.cpp:220:5 frame intel#17: libclang_rt.tsan_osx_dynamic.dylib`::__tsan_read8(void *) [inlined] __tsan::MemoryAccess(thr=0x00000001012fc000, pc=<unavailable>, addr=<unavailable>, size=8, typ=1) at tsan_rtl_access.cpp:442:3 frame intel#18: libclang_rt.tsan_osx_dynamic.dylib`::__tsan_read8(addr=<unavailable>) at tsan_interface.inc:34:3 <call into TSan from from instrumented code> thread intel#4, queue = 'com.apple.dock.fullscreen' frame #0: libsystem_kernel.dylib frame #1: libclang_rt.tsan_osx_dynamic.dylib`__sanitizer::FutexWait(p=<unavailable>, cmp=<unavailable>) at sanitizer_mac.cpp:540:3 frame #2: libclang_rt.tsan_osx_dynamic.dylib`__sanitizer::Semaphore::Wait(this=<unavailable>) at sanitizer_mutex.cpp:35:7 frame #3: libclang_rt.tsan_osx_dynamic.dylib`__sanitizer::Mutex::Lock(this=0x0000000102992a80) at sanitizer_mutex.h:196:18 frame intel#4: libclang_rt.tsan_osx_dynamic.dylib`__tsan::ScopedInterceptor::~ScopedInterceptor() [inlined] __sanitizer::GenericScopedLock<__sanitizer::Mutex>::GenericScopedLock(this=<unavailable>, mu=0x0000000102992a80) at sanitizer_mutex.h:383:10 frame intel#5: libclang_rt.tsan_osx_dynamic.dylib`__tsan::ScopedInterceptor::~ScopedInterceptor() [inlined] __sanitizer::GenericScopedLock<__sanitizer::Mutex>::GenericScopedLock(this=<unavailable>, mu=0x0000000102992a80) at sanitizer_mutex.h:382:77 frame intel#6: libclang_rt.tsan_osx_dynamic.dylib`__tsan::ScopedInterceptor::~ScopedInterceptor() at tsan_rtl.h:708:10 frame intel#7: libclang_rt.tsan_osx_dynamic.dylib`__tsan::ScopedInterceptor::~ScopedInterceptor() [inlined] __tsan::TryTraceFunc(thr=0x000000010f084000, pc=0) at tsan_rtl.h:751:7 frame intel#8: libclang_rt.tsan_osx_dynamic.dylib`__tsan::ScopedInterceptor::~ScopedInterceptor() [inlined] __tsan::FuncExit(thr=0x000000010f084000) at tsan_rtl.h:798:7 frame intel#9: libclang_rt.tsan_osx_dynamic.dylib`__tsan::ScopedInterceptor::~ScopedInterceptor(this=0x000000016f3ba280) at tsan_interceptors_posix.cpp:300:5 frame intel#10: libclang_rt.tsan_osx_dynamic.dylib`__tsan::ScopedInterceptor::~ScopedInterceptor(this=<unavailable>) at tsan_interceptors_posix.cpp:293:41 frame intel#11: libclang_rt.tsan_osx_dynamic.dylib`::wrap_os_unfair_lock_lock_with_options(lock=0x000000016f21b1e8, options=OS_UNFAIR_LOCK_NONE) at tsan_interceptors_mac.cpp:310:1 frame intel#12: dyld`dyld4::RuntimeLocks::withLoadersReadLock(this=0x000000016f21b1e0, work=0x00000001814525d4) block_pointer) at DyldRuntimeState.cpp:227:28 frame intel#13: dyld`dyld4::APIs::_dyld_get_image_vmaddr_slide(this=0x0000000101012a20, imageIndex=412) at DyldAPIs.cpp:273:11 frame intel#14: libclang_rt.tsan_osx_dynamic.dylib`__sanitizer::MemoryMappingLayout::Next(__sanitizer::MemoryMappedSegment*) at sanitizer_procmaps_mac.cpp:286:17 frame intel#15: libclang_rt.tsan_osx_dynamic.dylib`__sanitizer::MemoryMappingLayout::Next(this=0x000000016f3ba560, segment=0x000000016f3ba498) at sanitizer_procmaps_mac.cpp:432:15 frame intel#16: libclang_rt.tsan_osx_dynamic.dylib`__sanitizer::MemoryMappingLayout::DumpListOfModules(this=0x000000016f3ba560, modules=0x000000016f3ba618) at sanitizer_procmaps_mac.cpp:460:10 frame intel#17: libclang_rt.tsan_osx_dynamic.dylib`__sanitizer::ListOfModules::init(this=0x000000016f3ba618) at sanitizer_mac.cpp:610:18 frame intel#18: libclang_rt.tsan_osx_dynamic.dylib`__sanitizer::LibIgnore::OnLibraryLoaded(this=0x0000000101f3aa40, name="<some library>") at sanitizer_libignore.cpp:54:11 frame intel#19: libclang_rt.tsan_osx_dynamic.dylib`::wrap_dlopen(filename="<some library>", flag=<unavailable>) at sanitizer_common_interceptors.inc:6466:3 <library code> ``` rdar://106766395 Differential Revision: https://reviews.llvm.org/D146593

…callback The `TypeSystemMap::m_mutex` guards against concurrent modifications of members of `TypeSystemMap`. In particular, `m_map`. `TypeSystemMap::ForEach` iterates through the entire `m_map` calling a user-specified callback for each entry. This is all done while `m_mutex` is locked. However, there's nothing that guarantees that the callback itself won't call back into `TypeSystemMap` APIs on the same thread. This lead to double-locking `m_mutex`, which is undefined behaviour. We've seen this cause a deadlock in the swift plugin with following backtrace: ``` int main() { std::unique_ptr<int> up = std::make_unique<int>(5); volatile int val = *up; return val; } clang++ -std=c++2a -g -O1 main.cpp ./bin/lldb -o “br se -p return” -o run -o “v *up” -o “expr *up” -b ``` ``` frame intel#4: std::lock_guard<std::mutex>::lock_guard frame intel#5: lldb_private::TypeSystemMap::GetTypeSystemForLanguage <<<< Lock #2 frame intel#6: lldb_private::TypeSystemMap::GetTypeSystemForLanguage frame intel#7: lldb_private::Target::GetScratchTypeSystemForLanguage ... frame intel#26: lldb_private::SwiftASTContext::LoadLibraryUsingPaths frame intel#27: lldb_private::SwiftASTContext::LoadModule frame intel#30: swift::ModuleDecl::collectLinkLibraries frame intel#31: lldb_private::SwiftASTContext::LoadModule frame intel#34: lldb_private::SwiftASTContext::GetCompileUnitImportsImpl frame intel#35: lldb_private::SwiftASTContext::PerformCompileUnitImports frame intel#36: lldb_private::TypeSystemSwiftTypeRefForExpressions::GetSwiftASTContext frame intel#37: lldb_private::TypeSystemSwiftTypeRefForExpressions::GetPersistentExpressionState frame intel#38: lldb_private::Target::GetPersistentSymbol frame intel#41: lldb_private::TypeSystemMap::ForEach <<<< Lock #1 frame intel#42: lldb_private::Target::GetPersistentSymbol frame intel#43: lldb_private::IRExecutionUnit::FindInUserDefinedSymbols frame intel#44: lldb_private::IRExecutionUnit::FindSymbol frame intel#45: lldb_private::IRExecutionUnit::MemoryManager::GetSymbolAddressAndPresence frame intel#46: lldb_private::IRExecutionUnit::MemoryManager::findSymbol frame intel#47: non-virtual thunk to lldb_private::IRExecutionUnit::MemoryManager::findSymbol frame intel#48: llvm::LinkingSymbolResolver::findSymbol frame intel#49: llvm::LegacyJITSymbolResolver::lookup frame intel#50: llvm::RuntimeDyldImpl::resolveExternalSymbols frame intel#51: llvm::RuntimeDyldImpl::resolveRelocations frame intel#52: llvm::MCJIT::finalizeLoadedModules frame intel#53: llvm::MCJIT::finalizeObject frame intel#54: lldb_private::IRExecutionUnit::ReportAllocations frame intel#55: lldb_private::IRExecutionUnit::GetRunnableInfo frame intel#56: lldb_private::ClangExpressionParser::PrepareForExecution frame intel#57: lldb_private::ClangUserExpression::TryParse frame intel#58: lldb_private::ClangUserExpression::Parse ``` Our solution is to simply iterate over a local copy of `m_map`. **Testing** * Confirmed on manual reproducer (would reproduce 100% of the time before the patch) Differential Revision: https://reviews.llvm.org/D149949

…est unittest Need to finalize the DIBuilder to avoid leak sanitizer errors like this: Direct leak of 48 byte(s) in 1 object(s) allocated from: #0 0x55c99ea1761d in operator new(unsigned long) #1 0x55c9a518ae49 in operator new #2 0x55c9a518ae49 in llvm::MDTuple::getImpl(...) #3 0x55c9a4f1b1ec in getTemporary intel#4 0x55c9a4f1b1ec in llvm::DIBuilder::createFunction(...)

TSan reports the following data race: Write of size 4 at 0x000109e0b160 by thread T2 (mutexes: write M0, write M1): #0 NativeFile::Close() File.cpp:329 #1 ConnectionFileDescriptor::Disconnect(lldb_private::Status*) ConnectionFileDescriptorPosix.cpp:232 #2 Communication::Disconnect(lldb_private::Status*) Communication.cpp:61 #3 process_gdb_remote::ProcessGDBRemote::DidExit() ProcessGDBRemote.cpp:1164 intel#4 Process::SetExitStatus(int, char const*) Process.cpp:1097 intel#5 process_gdb_remote::ProcessGDBRemote::MonitorDebugserverProcess(...) ProcessGDBRemote.cpp:3387 Previous read of size 4 at 0x000109e0b160 by main thread (mutexes: write M2): #0 NativeFile::IsValid() const File.h:393 #1 ConnectionFileDescriptor::IsConnected() const ConnectionFileDescriptorPosix.cpp:121 #2 Communication::IsConnected() const Communication.cpp:79 #3 process_gdb_remote::GDBRemoteCommunication::WaitForPacketNoLock(...) GDBRemoteCommunication.cpp:256 intel#4 process_gdb_remote::GDBRemoteCommunication::WaitForPacketNoLock(...l) GDBRemoteCommunication.cpp:244 intel#5 process_gdb_remote::GDBRemoteClientBase::SendPacketAndWaitForResponseNoLock(llvm::StringRef, StringExtractorGDBRemote&) GDBRemoteClientBase.cpp:246 The problem is that in WaitForPacketNoLock's run loop, it checks that the connection is still connected. This races with the ConnectionFileDescriptor disconnecting. Most (but not all) access to the IOObject in ConnectionFileDescriptorPosix is already gated by the mutex. This patch just protects IsConnected in the same way. Differential revision: https://reviews.llvm.org/D157347

This reverts commit 0e63f1a. clang-format started to crash with contents like: a.h: ``` ``` $ clang-format a.h ``` PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace. Stack dump: 0. Program arguments: ../llvm/build/bin/clang-format a.h #0 0x0000560b689fe177 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) /usr/local/google/home/kadircet/repos/llvm/llvm/lib/Support/Unix/Signals.inc:723:13 #1 0x0000560b689fbfbe llvm::sys::RunSignalHandlers() /usr/local/google/home/kadircet/repos/llvm/llvm/lib/Support/Signals.cpp:106:18 #2 0x0000560b689feaca SignalHandler(int) /usr/local/google/home/kadircet/repos/llvm/llvm/lib/Support/Unix/Signals.inc:413:1 #3 0x00007f030405a540 (/lib/x86_64-linux-gnu/libc.so.6+0x3c540) intel#4 0x0000560b68a9a980 is /usr/local/google/home/kadircet/repos/llvm/clang/include/clang/Lex/Token.h:98:44 intel#5 0x0000560b68a9a980 is /usr/local/google/home/kadircet/repos/llvm/clang/lib/Format/FormatToken.h:562:51 intel#6 0x0000560b68a9a980 startsSequenceInternal<clang::tok::TokenKind, clang::tok::TokenKind> /usr/local/google/home/kadircet/repos/llvm/clang/lib/Format/FormatToken.h:831:9 intel#7 0x0000560b68a9a980 startsSequence<clang::tok::TokenKind, clang::tok::TokenKind> /usr/local/google/home/kadircet/repos/llvm/clang/lib/Format/FormatToken.h:600:12 intel#8 0x0000560b68a9a980 getFunctionName /usr/local/google/home/kadircet/repos/llvm/clang/lib/Format/TokenAnnotator.cpp:3131:17 intel#9 0x0000560b68a9a980 clang::format::TokenAnnotator::annotate(clang::format::AnnotatedLine&) /usr/local/google/home/kadircet/repos/llvm/clang/lib/Format/TokenAnnotator.cpp:3191:17 Segmentation fault ```

@andykaylor

This changes add SPIR-V translator support for the SPIR-V extension documented here: KhronosGroup/SPIRV-Registry#193. This extension adds one decoration to represent maximum error for FP operations and adds the related Capability. SPIRV Headers support for representing this in SPIR-V: KhronosGroup/SPIRV-Headers#363 intel#8134 added a new call-site attribute associated with FP builtin intrinsics. This attribute is named 'fpbuiltin-max-error'. Following example shows how this extension is supported in the translator. The input LLVM IR uses new LLVM builtin calls to represent FP operations. An attribute named 'fpbuiltin-max-error' is used to represent the max-error allowed in the FP operation. Example Input LLVM: %t6 = call float @llvm.fpbuiltin.sin.f32(float %f1) #2 attributes #2 = { "fpbuiltin-max-error"="2.5" } This is translated into a SPIR-V instruction (for add/sub/mul/div/rem) and OpenCl extended instruction for other instructions. A decoration to represent the max-error is attached to the SPIR-V instruction. SPIR-V code: 4 Decorate 97 FPMaxErrorDecorationINTEL 1075838976 6 ExtInst 2 97 1 sin 88 No new support is added to support translating this SPIR_V back to LLVM. Existing support is used. The decoration is translated back into named metadata associated with the LLVM instruction. This can be readily consumed by backends. Based on input from @andykaylor, we emit attributes when the FP operation is translated back to a call to a builtin function and emit metadata otherwise. Translated LLVM code for basic math functions (add/sub/mul/div/rem): %t6 = fmul float %f1, %f2, !fpbuiltin-max-error !7 !7 = !{!"2.500000"} Translated LLVM code for other math functions: %t6 = call spir_func float @_Z3sinf(float %f1) #3 attributes #3 = { "fpbuiltin-max-error"="4.000000" } Signed-off-by: Arvind Sudarsanam <arvind.sudarsanam@intel.com> Original commit: KhronosGroup/SPIRV-LLVM-Translator@c6fe12b

The new ACLE PR#225[1] now combines the slice parameters for some builtins. This patch is the #2 of 3 patches to update the interface. Slice specifies the ZA slice number directly and needs to be explicity implemented by the "user" with the base register plus the immediate offset [1]https://github.com/ARM-software/acle/pull/225/files

…fine.parallel verifier This patch updates AffineParallelOp::verify() to check each result type matches its corresponding reduction op (i.e, the result type must be a `FloatType` if the reduction attribute is `addf`) affine.parallel will crash on --lower-affine if the corresponding result type cannot match the reduction attribute. ``` %128 = affine.parallel (%arg2, %arg3) = (0, 0) to (8, 7) reduce ("maxf") -> (memref<8x7xf32>) { %alloc_33 = memref.alloc() : memref<8x7xf32> affine.yield %alloc_33 : memref<8x7xf32> } ``` This will crash and report a type conversion issue when we run `mlir-opt --lower-affine` ``` Assertion failed: (isa<To>(Val) && "cast<Ty>() argument of incompatible type!"), function cast, file Casting.h, line 572. PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace. Stack dump: 0. Program arguments: mlir-opt --lower-affine temp.mlir #0 0x0000000102a18f18 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/workspacebin/mlir-opt+0x1002f8f18) #1 0x0000000102a171b4 llvm::sys::RunSignalHandlers() (/workspacebin/mlir-opt+0x1002f71b4) #2 0x0000000102a195c4 SignalHandler(int) (/workspacebin/mlir-opt+0x1002f95c4) #3 0x00000001be7894c4 (/usr/lib/system/libsystem_platform.dylib+0x1803414c4) intel#4 0x00000001be771ee0 (/usr/lib/system/libsystem_pthread.dylib+0x180329ee0) intel#5 0x00000001be6ac340 (/usr/lib/system/libsystem_c.dylib+0x180264340) intel#6 0x00000001be6ab754 (/usr/lib/system/libsystem_c.dylib+0x180263754) intel#7 0x0000000106864790 mlir::arith::getIdentityValueAttr(mlir::arith::AtomicRMWKind, mlir::Type, mlir::OpBuilder&, mlir::Location) (.cold.4) (/workspacebin/mlir-opt+0x104144790) intel#8 0x0000000102ba66ac mlir::arith::getIdentityValueAttr(mlir::arith::AtomicRMWKind, mlir::Type, mlir::OpBuilder&, mlir::Location) (/workspacebin/mlir-opt+0x1004866ac) intel#9 0x0000000102ba6910 mlir::arith::getIdentityValue(mlir::arith::AtomicRMWKind, mlir::Type, mlir::OpBuilder&, mlir::Location) (/workspacebin/mlir-opt+0x100486910) ... ``` Fixes #64068 Reviewed By: mehdi_amini Differential Revision: https://reviews.llvm.org/D157985

…(#73463) Despite CWG2497 not being resolved, it is reasonable to expect the following code to compile (and which is supported by other compilers) ```cpp template<typename T> constexpr T f(); constexpr int g() { return f<int>(); } // #1 template<typename T> constexpr T f() { return 123; } int k[g()]; // #2 ``` To that end, we eagerly instantiate all referenced specializations of constexpr functions when they are defined. We maintain a map of (pattern, [instantiations]) independent of `PendingInstantiations` to avoid having to iterate that list after each function definition. We should apply the same logic to constexpr variables, but I wanted to keep the PR small. Fixes #73232

… on (#74207) lld string tail merging interacts badly with ASAN on Windows, as is reported in llvm/llvm-project#62078. A similar error was found when building LLVM with `-DLLVM_USE_SANITIZER=Address`: ```console [2/2] Building GenVT.inc... FAILED: include/llvm/CodeGen/GenVT.inc C:/Dev/llvm-project/Build_asan/include/llvm/CodeGen/GenVT.inc cmd.exe /C "cd /D C:\Dev\llvm-project\Build_asan && C:\Dev\llvm-project\Build_asan\bin\llvm-min-tblgen.exe -gen-vt -I C:/Dev/llvm-project/llvm/include/llvm/CodeGen -IC:/Dev/llvm-project/Build_asan/include -IC:/Dev/llvm-project/llvm/include C:/Dev/llvm-project/llvm/include/llvm/CodeGen/ValueTypes.td --write-if-changed -o include/llvm/CodeGen/GenVT.inc -d include/llvm/CodeGen/GenVT.inc.d" ================================================================= ==31944==ERROR: AddressSanitizer: global-buffer-overflow on address 0x7ff6cff80d20 at pc 0x7ff6cfcc7378 bp 0x00e8bcb8e990 sp 0x00e8bcb8e9d8 READ of size 1 at 0x7ff6cff80d20 thread T0 #0 0x7ff6cfcc7377 in strlen (C:\Dev\llvm-project\Build_asan\bin\llvm-min-tblgen.exe+0x1400a7377) #1 0x7ff6cfde50c2 in operator delete(void *, unsigned __int64) (C:\Dev\llvm-project\Build_asan\bin\llvm-min-tblgen.exe+0x1401c50c2) #2 0x7ff6cfdd75ef in operator delete(void *, unsigned __int64) (C:\Dev\llvm-project\Build_asan\bin\llvm-min-tblgen.exe+0x1401b75ef) #3 0x7ff6cfde59f9 in operator delete(void *, unsigned __int64) (C:\Dev\llvm-project\Build_asan\bin\llvm-min-tblgen.exe+0x1401c59f9) intel#4 0x7ff6cff03f6c in operator delete(void *, unsigned __int64) (C:\Dev\llvm-project\Build_asan\bin\llvm-min-tblgen.exe+0x1402e3f6c) intel#5 0x7ff6cfefbcbc in operator delete(void *, unsigned __int64) (C:\Dev\llvm-project\Build_asan\bin\llvm-min-tblgen.exe+0x1402dbcbc) intel#6 0x7ffb7f247343 (C:\WINDOWS\System32\KERNEL32.DLL+0x180017343) intel#7 0x7ffb800826b0 (C:\WINDOWS\SYSTEM32\ntdll.dll+0x1800526b0) 0x7ff6cff80d20 is located 31 bytes after global variable '"#error \"ArgKind is not defined\"\n"...' defined in 'C:\Dev\llvm-project\llvm\utils\TableGen\IntrinsicEmitter.cpp' (0x7ff6cff80ce0) of size 33 '"#error \"ArgKind is not defined\"\n"...' is ascii string '#error "ArgKind is not defined" ' 0x7ff6cff80d20 is located 0 bytes inside of global variable '""' defined in 'C:\Dev\llvm-project\llvm\utils\TableGen\IntrinsicEmitter.cpp' (0x7ff6cff80d20) of size 1 '""' is ascii string '' SUMMARY: AddressSanitizer: global-buffer-overflow (C:\Dev\llvm-project\Build_asan\bin\llvm-min-tblgen.exe+0x1400a7377) in strlen Shadow bytes around the buggy address: 0x7ff6cff80a80: 01 f9 f9 f9 f9 f9 f9 f9 00 00 00 00 01 f9 f9 f9 0x7ff6cff80b00: f9 f9 f9 f9 00 00 00 00 00 00 00 00 01 f9 f9 f9 0x7ff6cff80b80: f9 f9 f9 f9 00 00 00 00 01 f9 f9 f9 f9 f9 f9 f9 0x7ff6cff80c00: 00 00 00 00 01 f9 f9 f9 f9 f9 f9 f9 00 00 00 00 0x7ff6cff80c80: 00 00 00 00 01 f9 f9 f9 f9 f9 f9 f9 00 00 00 00 =>0x7ff6cff80d00: 01 f9 f9 f9[f9]f9 f9 f9 00 00 00 00 00 00 00 00 0x7ff6cff80d80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 0x7ff6cff80e00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 0x7ff6cff80e80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 0x7ff6cff80f00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 0x7ff6cff80f80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 Shadow byte legend (one shadow byte represents 8 application bytes): Addressable: 00 Partially addressable: 01 02 03 04 05 06 07 Heap left redzone: fa Freed heap region: fd Stack left redzone: f1 Stack mid redzone: f2 Stack right redzone: f3 Stack after return: f5 Stack use after scope: f8 Global redzone: f9 Global init order: f6 Poisoned by user: f7 Container overflow: fc Array cookie: ac Intra object redzone: bb ASan internal: fe Left alloca redzone: ca Right alloca redzone: cb ==31944==ABORTING ``` This is reproducible with the 17.0.3 release: ```console $ clang-cl --version clang version 17.0.3 Target: x86_64-pc-windows-msvc Thread model: posix InstalledDir: C:\Program Files\LLVM\bin $ cmake -S llvm -B Build -G Ninja -DLLVM_USE_SANITIZER=Address -DCMAKE_C_COMPILER=clang-cl -DCMAKE_CXX_COMPILER=clang-cl -DCMAKE_MSVC_RUNTIME_LIBRARY=MultiThreaded -DCMAKE_BUILD_TYPE=Release $ cd Build $ ninja all ```

…e defintion if available (#71004)" This reverts commit ef3feba. This caused an LLDB test failure on Linux for `lang/cpp/symbols/TestSymbols.test_dwo`: ``` make: Leaving directory '/home/worker/2.0.1/lldb-x86_64-debian/build/lldb-test-build.noindex/lang/cpp/symbols/TestSymbols.test_dwo' runCmd: expression -- D::i PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace. Stack dump: 0. HandleCommand(command = "expression -- D::i") 1. <user expression 0>:1:4: current parser token 'i' 2. <lldb wrapper prefix>:44:1: parsing function body '$__lldb_expr' 3. <lldb wrapper prefix>:44:1: in compound statement ('{}') Stack dump without symbol names (ensure you have llvm-symbolizer in your PATH or set the environment var `LLVM_SYMBOLIZER_PATH` to point to it): 0 _lldb.cpython-39-x86_64-linux-gnu.so 0x00007fbcfcb08b87 1 _lldb.cpython-39-x86_64-linux-gnu.so 0x00007fbcfcb067ae 2 _lldb.cpython-39-x86_64-linux-gnu.so 0x00007fbcfcb0923f 3 libpthread.so.0 0x00007fbd07ab7140 ``` And a failure in `TestCallStdStringFunction.py` on Linux aarch64: ``` -- Exit Code: -11 Command Output (stdout): -- lldb version 18.0.0git (https://github.com/llvm/llvm-project.git revision ef3feba) clang revision ef3feba llvm revision ef3feba -- Command Output (stderr): -- PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace. Stack dump: 0. HandleCommand(command = "expression str") 1. <lldb wrapper prefix>:45:34: current parser token ';' 2. <lldb wrapper prefix>:44:1: parsing function body '$__lldb_expr' 3. <lldb wrapper prefix>:44:1: in compound statement ('{}') #0 0x0000ffffb72a149c llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/tcwg-buildbot/worker/lldb-aarch64-ubuntu/build/lib/python3.8/site-packages/lldb/_[lldb.cpython-38-aarch64-linux-gnu.so](http://lldb.cpython-38-aarch64-linux-gnu.so/)+0x58c749c) #1 0x0000ffffb729f458 llvm::sys::RunSignalHandlers() (/home/tcwg-buildbot/worker/lldb-aarch64-ubuntu/build/lib/python3.8/site-packages/lldb/_[lldb.cpython-38-aarch64-linux-gnu.so](http://lldb.cpython-38-aarch64-linux-gnu.so/)+0x58c5458) #2 0x0000ffffb72a1bd0 SignalHandler(int) (/home/tcwg-buildbot/worker/lldb-aarch64-ubuntu/build/lib/python3.8/site-packages/lldb/_[lldb.cpython-38-aarch64-linux-gnu.so](http://lldb.cpython-38-aarch64-linux-gnu.so/)+0x58c7bd0) #3 0x0000ffffbdd9e7dc (linux-vdso.so.1+0x7dc) intel#4 0x0000ffffb71799d8 lldb_private::plugin::dwarf::SymbolFileDWARF::FindGlobalVariables(lldb_private::ConstString, lldb_private::CompilerDeclContext const&, unsigned int, lldb_private::VariableList&) (/home/tcwg-buildbot/worker/lldb-aarch64-ubuntu/build/lib/python3.8/site-packages/lldb/_[lldb.cpython-38-aarch64-linux-gnu.so](http://lldb.cpython-38-aarch64-linux-gnu.so/)+0x579f9d8) intel#5 0x0000ffffb7197508 DWARFASTParserClang::FindConstantOnVariableDefinition(lldb_private::plugin::dwarf::DWARFDIE) (/home/tcwg-buildbot/worker/lldb-aarch64-ubuntu/build/lib/python3.8/site-packages/lldb/_[lldb.cpython-38-aarch64-linux-gnu.so](http://lldb.cpython-38-aarch64-linux-gnu.so/)+0x57bd508) ```

The const.cpp testcase fails when running in MSVC mode, while it does succeed in MinGW mode. In MSVC mode, there are more constructor invocations than expected, as the printout looks like this: A(1), this = 0000025597930000 A(1), this = 0000025597930000 f: this = 0000025597930000, val = 1 A(1), this = 0000025597930000 f: this = 0000025597930000, val = 1 ~A, this = 0000025597930000, val = 1 ~A, this = 0000025597930000, val = 1 ~A, this = 0000025597930000, val = 1 While the expected printout looks like this: A(1), this = 000002C903E10000 f: this = 000002C903E10000, val = 1 f: this = 000002C903E10000, val = 1 ~A, this = 000002C903E10000, val = 1 Reapplying #70991 with the XFAIL changed to check the host triple, not the target triple. On an MSVC based build of Clang, but with the default target triple set to PS4/PS5, we will still see the failure. And a Linux based build of Clang that targets PS4/PS5 won't see the issue.

Internal builds of the unittests with msan flagged mempcpy_test. ==6862==WARNING: MemorySanitizer: use-of-uninitialized-value #0 0x55e34d7d734a in length llvm-project/libc/src/__support/CPP/string_view.h:41:11 #1 0x55e34d7d734a in string_view llvm-project/libc/src/__support/CPP/string_view.h:71:24 #2 0x55e34d7d734a in __llvm_libc_9999_0_0_git::testing::Test::testStrEq(char const*, char const*, char const*, char const*, __llvm_libc_9999_0_0_git::testing::internal::Location) llvm-project/libc/test/UnitTest/LibcTest.cpp:284:13 #3 0x55e34d7d4e09 in LlvmLibcMempcpyTest_Simple::Run() llvm-project/libc/test/src/string/mempcpy_test.cpp:20:3 intel#4 0x55e34d7d6dff in __llvm_libc_9999_0_0_git::testing::Test::runTests(char const*) llvm-project/libc/test/UnitTest/LibcTest.cpp:133:8 intel#5 0x55e34d7d86e0 in main llvm-project/libc/test/UnitTest/LibcTestMain.cpp:21:10 SUMMARY: MemorySanitizer: use-of-uninitialized-value llvm-project/libc/src/__support/CPP/string_view.h:41:11 in length What's going on here is that mempcpy_test.cpp's Simple test is using ASSERT_STREQ with a partially initialized char array. ASSERT_STREQ calls Test::testStrEq which constructs a cpp:string_view. That constructor calls the private method cpp::string_view::length. When built with msan, the loop is transformed into multi-byte access, which then fails upon access. I took a look at libc++'s __constexpr_strlen which just calls __builtin_strlen(). Replacing the implementation of cpp::string_view::length with a call to __builtin_strlen() may still result in out of bounds access when the test is built with msan. It's not safe to use ASSERT_STREQ with a partially initialized array. Initialize the whole array so that the test passes.

We'd like a way to select the current thread by its thread ID (rather than its internal LLDB thread index). This PR adds a `-t` option (`--thread_id` long option) that tells the `thread select` command to interpret the `<thread-index>` argument as a thread ID. Here's an example of it working: ``` michristensen@devbig356 llvm/llvm-project (thread-select-tid) » ../Debug/bin/lldb ~/scratch/cpp/threading/a.out (lldb) target create "/home/michristensen/scratch/cpp/threading/a.out" Current executable set to '/home/michristensen/scratch/cpp/threading/a.out' (x86_64). (lldb) b 18 Breakpoint 1: where = a.out`main + 80 at main.cpp:18:12, address = 0x0000000000000850 (lldb) run Process 215715 launched: '/home/michristensen/scratch/cpp/threading/a.out' (x86_64) This is a thread, i=1 This is a thread, i=2 This is a thread, i=3 This is a thread, i=4 This is a thread, i=5 Process 215715 stopped * thread #1, name = 'a.out', stop reason = breakpoint 1.1 frame #0: 0x0000555555400850 a.out`main at main.cpp:18:12 15 for (int i = 0; i < 5; i++) { 16 pthread_create(&thread_ids[i], NULL, foo, NULL); 17 } -> 18 for (int i = 0; i < 5; i++) { 19 pthread_join(thread_ids[i], NULL); 20 } 21 return 0; (lldb) thread select 2 * thread #2, name = 'a.out' frame #0: 0x00007ffff68f9918 libc.so.6`__nanosleep + 72 libc.so.6`__nanosleep: -> 0x7ffff68f9918 <+72>: cmpq $-0x1000, %rax ; imm = 0xF000 0x7ffff68f991e <+78>: ja 0x7ffff68f9952 ; <+130> 0x7ffff68f9920 <+80>: movl %edx, %edi 0x7ffff68f9922 <+82>: movl %eax, 0xc(%rsp) (lldb) thread info thread #2: tid = 216047, 0x00007ffff68f9918 libc.so.6`__nanosleep + 72, name = 'a.out' (lldb) thread list Process 215715 stopped thread #1: tid = 215715, 0x0000555555400850 a.out`main at main.cpp:18:12, name = 'a.out', stop reason = breakpoint 1.1 * thread #2: tid = 216047, 0x00007ffff68f9918 libc.so.6`__nanosleep + 72, name = 'a.out' thread #3: tid = 216048, 0x00007ffff68f9918 libc.so.6`__nanosleep + 72, name = 'a.out' thread intel#4: tid = 216049, 0x00007ffff68f9918 libc.so.6`__nanosleep + 72, name = 'a.out' thread intel#5: tid = 216050, 0x00007ffff68f9918 libc.so.6`__nanosleep + 72, name = 'a.out' thread intel#6: tid = 216051, 0x00007ffff68f9918 libc.so.6`__nanosleep + 72, name = 'a.out' (lldb) thread select 215715 error: invalid thread #215715. (lldb) thread select -t 215715 * thread #1, name = 'a.out', stop reason = breakpoint 1.1 frame #0: 0x0000555555400850 a.out`main at main.cpp:18:12 15 for (int i = 0; i < 5; i++) { 16 pthread_create(&thread_ids[i], NULL, foo, NULL); 17 } -> 18 for (int i = 0; i < 5; i++) { 19 pthread_join(thread_ids[i], NULL); 20 } 21 return 0; (lldb) thread select -t 216051 * thread intel#6, name = 'a.out' frame #0: 0x00007ffff68f9918 libc.so.6`__nanosleep + 72 libc.so.6`__nanosleep: -> 0x7ffff68f9918 <+72>: cmpq $-0x1000, %rax ; imm = 0xF000 0x7ffff68f991e <+78>: ja 0x7ffff68f9952 ; <+130> 0x7ffff68f9920 <+80>: movl %edx, %edi 0x7ffff68f9922 <+82>: movl %eax, 0xc(%rsp) (lldb) thread select 3 * thread #3, name = 'a.out' frame #0: 0x00007ffff68f9918 libc.so.6`__nanosleep + 72 libc.so.6`__nanosleep: -> 0x7ffff68f9918 <+72>: cmpq $-0x1000, %rax ; imm = 0xF000 0x7ffff68f991e <+78>: ja 0x7ffff68f9952 ; <+130> 0x7ffff68f9920 <+80>: movl %edx, %edi 0x7ffff68f9922 <+82>: movl %eax, 0xc(%rsp) (lldb) thread select -t 216048 * thread #3, name = 'a.out' frame #0: 0x00007ffff68f9918 libc.so.6`__nanosleep + 72 libc.so.6`__nanosleep: -> 0x7ffff68f9918 <+72>: cmpq $-0x1000, %rax ; imm = 0xF000 0x7ffff68f991e <+78>: ja 0x7ffff68f9952 ; <+130> 0x7ffff68f9920 <+80>: movl %edx, %edi 0x7ffff68f9922 <+82>: movl %eax, 0xc(%rsp) (lldb) thread select --thread_id 216048 * thread #3, name = 'a.out' frame #0: 0x00007ffff68f9918 libc.so.6`__nanosleep + 72 libc.so.6`__nanosleep: -> 0x7ffff68f9918 <+72>: cmpq $-0x1000, %rax ; imm = 0xF000 0x7ffff68f991e <+78>: ja 0x7ffff68f9952 ; <+130> 0x7ffff68f9920 <+80>: movl %edx, %edi 0x7ffff68f9922 <+82>: movl %eax, 0xc(%rsp) (lldb) help thread select Change the currently selected thread. Syntax: thread select <cmd-options> <thread-index> Command Options Usage: thread select [-t] <thread-index> -t ( --thread_id ) Provide a thread ID instead of a thread index. This command takes options and free-form arguments. If your arguments resemble option specifiers (i.e., they start with a - or --), you must use ' -- ' between the end of the command options and the beginning of the arguments. (lldb) c Process 215715 resuming Process 215715 exited with status = 0 (0x00000000) ```

Linalg op fusion (`Linalg/Transforms/Fusion.cpp`) used to generate invalid fused producer ops: ``` error: 'linalg.conv_2d_nhwc_hwcf' op expected type of operand #2 ('tensor<1x8x16x4xf32>') to match type of corresponding result ('tensor<?x?x?x?xf32>') note: see current operation: %24 = "linalg.conv_2d_nhwc_hwcf"(%21, %22, %23) <{dilations = dense<1> : tensor<2xi64>, operandSegmentSizes = array<i32: 2, 1>, strides = dense<2> : tensor<2xi64>}> ({ ^bb0(%arg9: f32, %arg10: f32, %arg11: f32): %28 = "arith.mulf"(%arg9, %arg10) <{fastmath = #arith.fastmath<none>}> : (f32, f32) -> f32 %29 = "arith.addf"(%arg11, %28) <{fastmath = #arith.fastmath<none>}> : (f32, f32) -> f32 "linalg.yield"(%29) : (f32) -> () }) {linalg.memoized_indexing_maps = [affine_map<(d0, d1, d2, d3, d4, d5, d6) -> (d0, d1 * 2 + d4, d2 * 2 + d5, d6)>, affine_map<(d0, d1, d2, d3, d4, d5, d6) -> (d4, d5, d6, d3)>, affine_map<(d0, d1, d2, d3, d4, d5, d6) -> (d0, d1, d2, d3)>]} : (tensor<1x?x?x3xf32>, tensor<3x3x3x4xf32>, tensor<1x8x16x4xf32>) -> tensor<?x?x?x?xf32> ``` This is a problem because the input IR to greedy pattern rewriter during `-test-linalg-greedy-fusion` is invalid. This commit fixes tests such as `mlir/test/Dialect/Linalg/tile-and-fuse-tensors.mlir` when verifying the IR after each pattern application (#74270).

This has been flaky for a while, for example https://lab.llvm.org/buildbot/#/builders/96/builds/50350 ``` Command Output (stdout): -- lldb version 18.0.0git (https://github.com/llvm/llvm-project.git revision 3974d89) clang revision 3974d89 llvm revision 3974d89 "can't evaluate expressions when the process is running." ``` ``` PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace. #0 0x0000ffffa46191a0 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/home/tcwg-buildbot/worker/lldb-aarch64-ubuntu/build/lib/python3.8/site-packages/lldb/_lldb.cpython-38-aarch64-linux-gnu.so+0x529a1a0) #1 0x0000ffffa4617144 llvm::sys::RunSignalHandlers() (/home/tcwg-buildbot/worker/lldb-aarch64-ubuntu/build/lib/python3.8/site-packages/lldb/_lldb.cpython-38-aarch64-linux-gnu.so+0x5298144) #2 0x0000ffffa46198d0 SignalHandler(int) (/home/tcwg-buildbot/worker/lldb-aarch64-ubuntu/build/lib/python3.8/site-packages/lldb/_lldb.cpython-38-aarch64-linux-gnu.so+0x529a8d0) #3 0x0000ffffab25b7dc (linux-vdso.so.1+0x7dc) intel#4 0x0000ffffab13d050 /build/glibc-Q8DG8B/glibc-2.31/string/../sysdeps/aarch64/multiarch/memcpy_advsimd.S:92:0 intel#5 0x0000ffffa446f420 lldb_private::process_gdb_remote::GDBRemoteRegisterContext::PrivateSetRegisterValue(unsigned int, llvm::ArrayRef<unsigned char>) (/home/tcwg-buildbot/worker/lldb-aarch64-ubuntu/build/lib/python3.8/site-packages/lldb/_lldb.cpython-38-aarch64-linux-gnu.so+0x50f0420) intel#6 0x0000ffffa446f7b8 lldb_private::process_gdb_remote::GDBRemoteRegisterContext::GetPrimordialRegister(lldb_private::RegisterInfo const*, lldb_private::process_gdb_remote::GDBRemoteCommunicationClient&) (/home/tcwg-buildbot/worker/lldb-aarch64-ubuntu/build/lib/python3.8/site-packages/lldb/_lldb.cpython-38-aarch64-linux-gnu.so+0x50f07b8) intel#7 0x0000ffffa446f308 lldb_private::process_gdb_remote::GDBRemoteRegisterContext::ReadRegisterBytes(lldb_private::RegisterInfo const*) (/home/tcwg-buildbot/worker/lldb-aarch64-ubuntu/build/lib/python3.8/site-packages/lldb/_lldb.cpython-38-aarch64-linux-gnu.so+0x50f0308) intel#8 0x0000ffffa446ec1c lldb_private::process_gdb_remote::GDBRemoteRegisterContext::ReadRegister(lldb_private::RegisterInfo const*, lldb_private::RegisterValue&) (/home/tcwg-buildbot/worker/lldb-aarch64-ubuntu/build/lib/python3.8/site-packages/lldb/_lldb.cpython-38-aarch64-linux-gnu.so+0x50efc1c) intel#9 0x0000ffffa412eaa4 lldb_private::RegisterContext::ReadRegisterAsUnsigned(lldb_private::RegisterInfo const*, unsigned long) (/home/tcwg-buildbot/worker/lldb-aarch64-ubuntu/build/lib/python3.8/site-packages/lldb/_lldb.cpython-38-aarch64-linux-gnu.so+0x4dafaa4) intel#10 0x0000ffffa420861c ReadLinuxProcessAddressMask(std::shared_ptr<lldb_private::Process>, llvm::StringRef) (/home/tcwg-buildbot/worker/lldb-aarch64-ubuntu/build/lib/python3.8/site-packages/lldb/_lldb.cpython-38-aarch64-linux-gnu.so+0x4e8961c) intel#11 0x0000ffffa4208430 ABISysV_arm64::FixCodeAddress(unsigned long) (/home/tcwg-buildbot/worker/lldb-aarch64-ubuntu/build/lib/python3.8/site-packages/lldb/_lldb.cpython-38-aarch64-linux-gnu.so+0x4e89430) ``` Judging by the backtrace something is trying to read the pointer authentication address/code mask registers. This explains why I've not seen this issue locally, as the buildbot runs on Graviton 3 with has the pointer authentication extension. I will try to reproduce, fix and re-enable the test.

…5394) Calling one of pthread join/detach interceptor on an already joined/detached thread causes asserts such as: AddressSanitizer: CHECK failed: sanitizer_thread_arg_retval.cpp:56 "((t)) != (0)" (0x0, 0x0) (tid=1236094) #0 0x555555634f8b in __asan::CheckUnwind() compiler-rt/lib/asan/asan_rtl.cpp:69:3 #1 0x55555564e06e in __sanitizer::CheckFailed(char const*, int, char const*, unsigned long long, unsigned long long) compiler-rt/lib/sanitizer_common/sanitizer_termination.cpp:86:24 #2 0x5555556491df in __sanitizer::ThreadArgRetval::BeforeJoin(unsigned long) const compiler-rt/lib/sanitizer_common/sanitizer_thread_arg_retval.cpp:56:3 #3 0x5555556198ed in Join<___interceptor_pthread_tryjoin_np(void*, void**)::<lambda()> > compiler-rt/lib/asan/../sanitizer_common/sanitizer_thread_arg_retval.h:74:26 intel#4 0x5555556198ed in pthread_tryjoin_np compiler-rt/lib/asan/asan_interceptors.cpp:311:29 The assert are replaced by error codes.

Add custom combine to lower load <3 x i8> as the more efficient sequence below: ldrb wX, [x0, #2] ldrh wY, [x0] orr wX, wY, wX, lsl intel#16 fmov s0, wX At the moment, there are almost no cases in which such vector operations will be generated automatically. The motivating case is non-power-of-2 SLP vectorization: llvm/llvm-project#77790

TestCases/Misc/Linux/sigaction.cpp fails because dlsym() may call malloc on failure. And then the wrapped malloc appears to access thread local storage using global dynamic accesses, thus calling ___interceptor___tls_get_addr, before REAL(__tls_get_addr) has been set, so we get a crash inside ___interceptor___tls_get_addr. For example, this can happen when looking up __isoc23_scanf which might not exist in some libcs. Fix this by marking the thread local variable accessed inside the debug checks as "initial-exec", which does not require __tls_get_addr. This is probably a better alternative to llvm/llvm-project#83886. This fixes a different crash but is related to llvm/llvm-project#46204. Backtrace: ``` #0 0x0000000000000000 in ?? () #1 0x00007ffff6a9d89e in ___interceptor___tls_get_addr (arg=0x7ffff6b27be8) at /path/to/llvm/compiler-rt/lib/tsan/rtl/tsan_interceptors_posix.cpp:2759 #2 0x00007ffff6a46bc6 in __sanitizer::CheckedMutex::LockImpl (this=0x7ffff6b27be8, pc=140737331846066) at /path/to/llvm/compiler-rt/lib/sanitizer_common/sanitizer_mutex.cpp:218 #3 0x00007ffff6a448b2 in __sanitizer::CheckedMutex::Lock (this=0x7ffff6b27be8, this@entry=0x730000000580) at /path/to/llvm/compiler-rt/lib/tsan/rtl/../../sanitizer_common/sanitizer_mutex.h:129 intel#4 __sanitizer::Mutex::Lock (this=0x7ffff6b27be8, this@entry=0x730000000580) at /path/to/llvm/compiler-rt/lib/tsan/rtl/../../sanitizer_common/sanitizer_mutex.h:167 intel#5 0x00007ffff6abdbb2 in __sanitizer::GenericScopedLock<__sanitizer::Mutex>::GenericScopedLock (mu=0x730000000580, this=<optimized out>) at /path/to/llvm/compiler-rt/lib/tsan/rtl/../../sanitizer_common/sanitizer_mutex.h:383 intel#6 __sanitizer::SizeClassAllocator64<__tsan::AP64>::GetFromAllocator (this=0x7ffff7487dc0 <__tsan::allocator_placeholder>, stat=stat@entry=0x7ffff570db68, class_id=11, chunks=chunks@entry=0x7ffff5702cc8, n_chunks=n_chunks@entry=128) at /path/to/llvm/compiler-rt/lib/tsan/rtl/../../sanitizer_common/sanitizer_allocator_primary64.h:207 intel#7 0x00007ffff6abdaa0 in __sanitizer::SizeClassAllocator64LocalCache<__sanitizer::SizeClassAllocator64<__tsan::AP64> >::Refill (this=<optimized out>, c=c@entry=0x7ffff5702cb8, allocator=<optimized out>, class_id=<optimized out>) at /path/to/llvm/compiler-rt/lib/tsan/rtl/../../sanitizer_common/sanitizer_allocator_local_cache.h:103 intel#8 0x00007ffff6abd731 in __sanitizer::SizeClassAllocator64LocalCache<__sanitizer::SizeClassAllocator64<__tsan::AP64> >::Allocate (this=0x7ffff6b27be8, allocator=0x7ffff5702cc8, class_id=140737311157448) at /path/to/llvm/compiler-rt/lib/tsan/rtl/../../sanitizer_common/sanitizer_allocator_local_cache.h:39 intel#9 0x00007ffff6abc397 in __sanitizer::CombinedAllocator<__sanitizer::SizeClassAllocator64<__tsan::AP64>, __sanitizer::LargeMmapAllocatorPtrArrayDynamic>::Allocate (this=0x7ffff5702cc8, cache=0x7ffff6b27be8, size=<optimized out>, size@entry=175, alignment=alignment@entry=16) at /path/to/llvm/compiler-rt/lib/tsan/rtl/../../sanitizer_common/sanitizer_allocator_combined.h:69 intel#10 0x00007ffff6abaa6a in __tsan::user_alloc_internal (thr=0x7ffff7ebd980, pc=140737331499943, sz=sz@entry=175, align=align@entry=16, signal=true) at /path/to/llvm/compiler-rt/lib/tsan/rtl/tsan_mman.cpp:198 intel#11 0x00007ffff6abb0d1 in __tsan::user_alloc (thr=0x7ffff6b27be8, pc=140737331846066, sz=11, sz@entry=175) at /path/to/llvm/compiler-rt/lib/tsan/rtl/tsan_mman.cpp:223 intel#12 0x00007ffff6a693b5 in ___interceptor_malloc (size=175) at /path/to/llvm/compiler-rt/lib/tsan/rtl/tsan_interceptors_posix.cpp:666 intel#13 0x00007ffff7fce7f2 in malloc (size=175) at ../include/rtld-malloc.h:56 intel#14 __GI__dl_exception_create_format (exception=exception@entry=0x7fffffffd0d0, objname=0x7ffff7fc3550 "/path/to/llvm/compiler-rt/cmake-build-all-sanitizers/lib/linux/libclang_rt.tsan-x86_64.so", fmt=fmt@entry=0x7ffff7ff2db9 "undefined symbol: %s%s%s") at ./elf/dl-exception.c:157 intel#15 0x00007ffff7fd50e8 in _dl_lookup_symbol_x (undef_name=0x7ffff6af868b "__isoc23_scanf", undef_map=<optimized out>, ref=0x7fffffffd148, symbol_scope=<optimized out>, version=<optimized out>, type_class=0, flags=2, skip_map=0x7ffff7fc35e0) at ./elf/dl-lookup.c:793 --Type <RET> for more, q to quit, c to continue without paging-- intel#16 0x00007ffff656d6ed in do_sym (handle=<optimized out>, name=0x7ffff6af868b "__isoc23_scanf", who=0x7ffff6a3bb84 <__interception::InterceptFunction(char const*, unsigned long*, unsigned long, unsigned long)+36>, vers=vers@entry=0x0, flags=flags@entry=2) at ./elf/dl-sym.c:146 intel#17 0x00007ffff656d9dd in _dl_sym (handle=<optimized out>, name=<optimized out>, who=<optimized out>) at ./elf/dl-sym.c:195 intel#18 0x00007ffff64a2854 in dlsym_doit (a=a@entry=0x7fffffffd3b0) at ./dlfcn/dlsym.c:40 intel#19 0x00007ffff7fcc489 in __GI__dl_catch_exception (exception=exception@entry=0x7fffffffd310, operate=0x7ffff64a2840 <dlsym_doit>, args=0x7fffffffd3b0) at ./elf/dl-catch.c:237 intel#20 0x00007ffff7fcc5af in _dl_catch_error (objname=0x7fffffffd368, errstring=0x7fffffffd370, mallocedp=0x7fffffffd367, operate=<optimized out>, args=<optimized out>) at ./elf/dl-catch.c:256 intel#21 0x00007ffff64a2257 in _dlerror_run (operate=operate@entry=0x7ffff64a2840 <dlsym_doit>, args=args@entry=0x7fffffffd3b0) at ./dlfcn/dlerror.c:138 intel#22 0x00007ffff64a28e5 in dlsym_implementation (dl_caller=<optimized out>, name=<optimized out>, handle=<optimized out>) at ./dlfcn/dlsym.c:54 intel#23 ___dlsym (handle=<optimized out>, name=<optimized out>) at ./dlfcn/dlsym.c:68 intel#24 0x00007ffff6a3bb84 in __interception::GetFuncAddr (name=0x7ffff6af868b "__isoc23_scanf", trampoline=140737311157448) at /path/to/llvm/compiler-rt/lib/interception/interception_linux.cpp:42 intel#25 __interception::InterceptFunction (name=0x7ffff6af868b "__isoc23_scanf", ptr_to_real=0x7ffff74850e8 <__interception::real___isoc23_scanf>, func=11, trampoline=140737311157448) at /path/to/llvm/compiler-rt/lib/interception/interception_linux.cpp:61 intel#26 0x00007ffff6a9f2d9 in InitializeCommonInterceptors () at /path/to/llvm/compiler-rt/lib/tsan/rtl/../../sanitizer_common/sanitizer_common_interceptors.inc:10315 ``` Reviewed By: vitalybuka, MaskRay Pull Request: llvm/llvm-project#83890

…0820) This solves some ambuguity introduced in P0522 regarding how template template parameters are partially ordered, and should reduce the negative impact of enabling `-frelaxed-template-template-args` by default. When performing template argument deduction, a template template parameter containing no packs should be more specialized than one that does. Given the following example: ```C++ template<class T2> struct A; template<template<class ...T3s> class TT1, class T4> struct A<TT1<T4>>; // #1 template<template<class T5 > class TT2, class T6> struct A<TT2<T6>>; // #2 template<class T1> struct B; template struct A<B<char>>; ``` Prior to P0522, candidate `#2` would be more specialized. After P0522, neither is more specialized, so this becomes ambiguous. With this change, `#2` becomes more specialized again, maintaining compatibility with pre-P0522 implementations. The problem is that in P0522, candidates are at least as specialized when matching packs to fixed-size lists both ways, whereas before, a fixed-size list is more specialized. This patch keeps the original behavior when checking template arguments outside deduction, but restores this aspect of pre-P0522 matching during deduction. --- Since this changes provisional implementation of CWG2398 which has not been released yet, and already contains a changelog entry, we don't provide a changelog entry here.

'reduction' has a few restrictions over normal 'var-list' clauses: 1- On parallel, a num_gangs can only have 1 argument when combined with reduction. These two aren't able to be combined on any other of the compute constructs however. 2- The vars all must be 'numerical data types' types of some sort, or a 'composite of numerical data types'. A list of types is given in the standard as a minimum, so we choose 'isScalar', which covers all of these types and keeps types that are actually numeric. Other compilers don't seem to implement the 'composite of numerical data types', though we do. 3- Because of the above restrictions, member-of-composite is not allowed, so any access via a memberexpr is disallowed. Array-element and sub-arrays (aka array sections) are both permitted, so long as they meet the requirements of #2. This patch implements all of these for compute constructs.

… (#92855) This solves some ambuguity introduced in P0522 regarding how template template parameters are partially ordered, and should reduce the negative impact of enabling `-frelaxed-template-template-args` by default. When performing template argument deduction, we extend the provisional wording introduced in llvm/llvm-project#89807 so it also covers deduction of class templates. Given the following example: ```C++ template <class T1, class T2 = float> struct A; template <class T3> struct B; template <template <class T4> class TT1, class T5> struct B<TT1<T5>>; // #1 template <class T6, class T7> struct B<A<T6, T7>>; // #2 template struct B<A<int>>; ``` Prior to P0522, `#2` was picked. Afterwards, this became ambiguous. This patch restores the pre-P0522 behavior, `#2` is picked again. This has the beneficial side effect of making the following code valid: ```C++ template<class T, class U> struct A {}; A<int, float> v; template<template<class> class TT> void f(TT<int>); // OK: TT picks 'float' as the default argument for the second parameter. void g() { f(v); } ``` --- Since this changes provisional implementation of CWG2398 which has not been released yet, and already contains a changelog entry, we don't provide a changelog entry here.

sommerlukas mentioned this pull request Feb 15, 2023

Kernel Fusion support for CUDA backend sommerlukas/llvm-test-suite#3

Closed

sommerlukas had a problem deploying to aws February 17, 2023 11:21 — with GitHub Actions Failure

victor-eds reviewed Mar 6, 2023

View reviewed changes

sommerlukas force-pushed the experiments/cuda-fusion branch 2 times, most recently from 57df68b to 5200cf6 Compare March 7, 2023 14:37

sommerlukas had a problem deploying to aws March 9, 2023 11:21 — with GitHub Actions Failure

victor-eds reviewed Mar 9, 2023

View reviewed changes

sycl-fusion/passes/target/TargetFusionInfo.cpp Outdated Show resolved Hide resolved

sycl-fusion/passes/target/TargetFusionInfo.cpp Outdated Show resolved Hide resolved

sommerlukas self-assigned this Mar 10, 2023

sommerlukas had a problem deploying to aws March 12, 2023 11:21 — with GitHub Actions Failure

victor-eds approved these changes Mar 13, 2023

View reviewed changes

Naghasan reviewed Mar 14, 2023

View reviewed changes

sycl-fusion/jit-compiler/lib/translation/KernelTranslation.cpp Outdated Show resolved Hide resolved

sycl-fusion/jit-compiler/lib/translation/KernelTranslation.cpp Outdated Show resolved Hide resolved

sycl/source/detail/scheduler/commands.hpp Show resolved Hide resolved

sommerlukas force-pushed the experiments/cuda-fusion branch 2 times, most recently from 56cbae1 to 5de0b84 Compare March 22, 2023 16:21

sommerlukas requested review from victor-eds and Naghasan and removed request for victor-eds and Naghasan March 22, 2023 16:48

Naghasan approved these changes Mar 22, 2023

View reviewed changes

sommerlukas added 8 commits March 23, 2023 07:50

[SYCL][Fusion] Embed LLVM IR for SYCL for Nvidia

cd2497c

Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

Enable LLVM IR as alternative fusion input format;

57bd8f6

Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

[SYCL][Fusion] Add result translation to PTX

0244fc0

Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

[SYCL][Fusion] Provide correct target spec;

7098094

Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

[SYCL][Fusion] Avoid removing dependencies

2b8e92d

Avoid the dependencies of the dependencies to be removed when cleaning up the input commands from the graph without executing them. Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

[SYCL][Fusion] Set device binary image format

fc43285

Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

[SYCL][Fusion] Refactor target-specific processing

6fad170

Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

[SYCL][Fusion] Do not require null terminator

ce6855a

Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

sommerlukas added 4 commits March 23, 2023 07:50

[SYCL][Fusion] Enable JIT caching for CUDA fusion

4506db6

Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

[SYCL][Fusion] Catch empty standard arguments

26978b2

Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

[SYCL][Fusion] Rebase and address feedback

aa2b629

Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

[SYCL][Fusion] Update linkage graph diagram

d323beb

Signed-off-by: Lukas Sommer <lukas.sommer@codeplay.com>

sommerlukas force-pushed the experiments/cuda-fusion branch from 585e098 to d323beb Compare March 23, 2023 08:36

sommerlukas closed this Mar 23, 2023

sommerlukas had a problem deploying to aws March 24, 2023 11:21 — with GitHub Actions Failure

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Kernel Fusion support for CUDA backend #2

Kernel Fusion support for CUDA backend #2

sommerlukas commented Feb 15, 2023

sommerlukas commented Mar 23, 2023

Kernel Fusion support for CUDA backend #2

Kernel Fusion support for CUDA backend #2

Conversation

sommerlukas commented Feb 15, 2023

sommerlukas commented Mar 23, 2023