Author: Ethan Luis McDonough Date: 2025-06-10T20:19:38-05:00 New Revision: 67ff66e67734c0b283ec676899e5b89b67fdafcb
URL: https://github.com/llvm/llvm-project/commit/67ff66e67734c0b283ec676899e5b89b67fdafcb DIFF: https://github.com/llvm/llvm-project/commit/67ff66e67734c0b283ec676899e5b89b67fdafcb.diff LOG: [PGO][Offload] Fix offload coverage mapping (#143490) This pull request fixes coverage mapping on GPU targets. - It adds an address space cast to the coverage mapping generation pass. - It reads the profiled function names from the ELF directly. Reading it from public globals was causing issues in cases where multiple device-code object files are linked together. Added: Modified: clang/lib/CodeGen/CoverageMappingGen.cpp llvm/lib/Transforms/Instrumentation/InstrProfiling.cpp offload/plugins-nextgen/common/include/GlobalHandler.h offload/plugins-nextgen/common/src/GlobalHandler.cpp offload/plugins-nextgen/common/src/PluginInterface.cpp Removed: ################################################################################ diff --git a/clang/lib/CodeGen/CoverageMappingGen.cpp b/clang/lib/CodeGen/CoverageMappingGen.cpp index 1788bb4f28697..4aafac349e3e9 100644 --- a/clang/lib/CodeGen/CoverageMappingGen.cpp +++ b/clang/lib/CodeGen/CoverageMappingGen.cpp @@ -2622,8 +2622,9 @@ void CoverageMappingModuleGen::emit() { CGM.addUsedGlobal(CovData); // Create the deferred function records array if (!FunctionNames.empty()) { - auto NamesArrTy = llvm::ArrayType::get(llvm::PointerType::getUnqual(Ctx), - FunctionNames.size()); + auto AddrSpace = FunctionNames.front()->getType()->getPointerAddressSpace(); + auto NamesArrTy = llvm::ArrayType::get( + llvm::PointerType::get(Ctx, AddrSpace), FunctionNames.size()); auto NamesArrVal = llvm::ConstantArray::get(NamesArrTy, FunctionNames); // This variable will *NOT* be emitted to the object file. It is used // to pass the list of names referenced to codegen. diff --git a/llvm/lib/Transforms/Instrumentation/InstrProfiling.cpp b/llvm/lib/Transforms/Instrumentation/InstrProfiling.cpp index fe3b0da33a009..5e7548b0a2fd1 100644 --- a/llvm/lib/Transforms/Instrumentation/InstrProfiling.cpp +++ b/llvm/lib/Transforms/Instrumentation/InstrProfiling.cpp @@ -1955,12 +1955,6 @@ void InstrLowerer::emitNameData() { GlobalValue::PrivateLinkage, NamesVal, getInstrProfNamesVarName()); - // Make names variable public if current target is a GPU - if (isGPUProfTarget(M)) { - NamesVar->setLinkage(GlobalValue::ExternalLinkage); - NamesVar->setVisibility(GlobalValue::VisibilityTypes::ProtectedVisibility); - } - NamesSize = CompressedNameStr.size(); setGlobalVariableLargeSection(TT, *NamesVar); NamesVar->setSection( diff --git a/offload/plugins-nextgen/common/include/GlobalHandler.h b/offload/plugins-nextgen/common/include/GlobalHandler.h index 6def53430a7c0..5d6109df49da5 100644 --- a/offload/plugins-nextgen/common/include/GlobalHandler.h +++ b/offload/plugins-nextgen/common/include/GlobalHandler.h @@ -80,6 +80,7 @@ struct GPUProfGlobals { void dump() const; Error write() const; + bool empty() const; }; /// Subclass of GlobalTy that holds the memory for a global of \p Ty. @@ -192,9 +193,6 @@ class GenericGlobalHandlerTy { /*D2H=*/false); } - /// Checks whether a given image contains profiling globals. - bool hasProfilingGlobals(GenericDeviceTy &Device, DeviceImageTy &Image); - /// Reads profiling data from a GPU image to supplied profdata struct. /// Iterates through the image symbol table and stores global values /// with profiling prefixes. diff --git a/offload/plugins-nextgen/common/src/GlobalHandler.cpp b/offload/plugins-nextgen/common/src/GlobalHandler.cpp index 27d7e8ee2fdf3..5464c197dba78 100644 --- a/offload/plugins-nextgen/common/src/GlobalHandler.cpp +++ b/offload/plugins-nextgen/common/src/GlobalHandler.cpp @@ -173,16 +173,6 @@ Error GenericGlobalHandlerTy::readGlobalFromImage(GenericDeviceTy &Device, return Plugin::success(); } -bool GenericGlobalHandlerTy::hasProfilingGlobals(GenericDeviceTy &Device, - DeviceImageTy &Image) { - GlobalTy global(getInstrProfNamesVarName().str(), 0); - if (auto Err = getGlobalMetadataFromImage(Device, Image, global)) { - consumeError(std::move(Err)); - return false; - } - return true; -} - Expected<GPUProfGlobals> GenericGlobalHandlerTy::readProfilingGlobals(GenericDeviceTy &Device, DeviceImageTy &Image) { @@ -204,12 +194,17 @@ GenericGlobalHandlerTy::readProfilingGlobals(GenericDeviceTy &Device, // Check if given current global is a profiling global based // on name if (*NameOrErr == getInstrProfNamesVarName()) { - // Read in profiled function names - DeviceProfileData.NamesData = SmallVector<uint8_t>(Sym.getSize(), 0); - GlobalTy NamesGlobal(NameOrErr->str(), Sym.getSize(), - DeviceProfileData.NamesData.data()); - if (auto Err = readGlobalFromDevice(Device, Image, NamesGlobal)) - return Err; + // Read in profiled function names from ELF + auto SectionOrErr = Sym.getSection(); + if (!SectionOrErr) + return SectionOrErr.takeError(); + + auto ContentsOrErr = (*SectionOrErr)->getContents(); + if (!ContentsOrErr) + return ContentsOrErr.takeError(); + + SmallVector<uint8_t> NameBytes(ContentsOrErr->bytes()); + DeviceProfileData.NamesData = NameBytes; } else if (NameOrErr->starts_with(getInstrProfCountersVarPrefix())) { // Read global variable profiling counts SmallVector<int64_t> Counts(Sym.getSize() / sizeof(int64_t), 0); @@ -322,3 +317,7 @@ Error GPUProfGlobals::write() const { return Plugin::success(); } + +bool GPUProfGlobals::empty() const { + return Counts.empty() && Data.empty() && NamesData.empty(); +} diff --git a/offload/plugins-nextgen/common/src/PluginInterface.cpp b/offload/plugins-nextgen/common/src/PluginInterface.cpp index f9e316adad8f4..f9a6b3c1f4324 100644 --- a/offload/plugins-nextgen/common/src/PluginInterface.cpp +++ b/offload/plugins-nextgen/common/src/PluginInterface.cpp @@ -858,14 +858,13 @@ Error GenericDeviceTy::deinit(GenericPluginTy &Plugin) { for (auto *Image : LoadedImages) { GenericGlobalHandlerTy &Handler = Plugin.getGlobalHandler(); - if (!Handler.hasProfilingGlobals(*this, *Image)) - continue; - - GPUProfGlobals profdata; auto ProfOrErr = Handler.readProfilingGlobals(*this, *Image); if (!ProfOrErr) return ProfOrErr.takeError(); + if (ProfOrErr->empty()) + continue; + // Dump out profdata if ((OMPX_DebugKind.get() & uint32_t(DeviceDebugKind::PGODump)) == uint32_t(DeviceDebugKind::PGODump)) _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
