https://github.com/anutosh491 updated
https://github.com/llvm/llvm-project/pull/136404
>From 1f7205615f8d11c1b58e2a2760f85663f97767c5 Mon Sep 17 00:00:00 2001
From: anutosh491 <andersonbhat...@gmail.com>
Date: Fri, 18 Apr 2025 18:45:00 +0530
Subject: [PATCH 1/3] Fix cuda flag with clang-repl
---
clang/include/clang/Interpreter/Interpreter.h | 13 ++-
clang/lib/Interpreter/DeviceOffload.cpp | 43 +++++----
clang/lib/Interpreter/DeviceOffload.h | 4 +-
clang/lib/Interpreter/Interpreter.cpp | 89 ++++++++++++-------
4 files changed, 88 insertions(+), 61 deletions(-)
diff --git a/clang/include/clang/Interpreter/Interpreter.h
b/clang/include/clang/Interpreter/Interpreter.h
index b1b63aedf86ab..1b228e0917d02 100644
--- a/clang/include/clang/Interpreter/Interpreter.h
+++ b/clang/include/clang/Interpreter/Interpreter.h
@@ -95,6 +95,9 @@ class Interpreter {
// An optional parser for CUDA offloading
std::unique_ptr<IncrementalParser> DeviceParser;
+ // An optional action for CUDA offloading
+ std::unique_ptr<IncrementalAction> DeviceAct;
+
/// List containing information about each incrementally parsed piece of
code.
std::list<PartialTranslationUnit> PTUs;
@@ -129,7 +132,8 @@ class Interpreter {
public:
virtual ~Interpreter();
static llvm::Expected<std::unique_ptr<Interpreter>>
- create(std::unique_ptr<CompilerInstance> CI);
+ create(std::unique_ptr<CompilerInstance> CI,
+ std::unique_ptr<CompilerInstance> DeviceCI = nullptr);
static llvm::Expected<std::unique_ptr<Interpreter>>
createWithCUDA(std::unique_ptr<CompilerInstance> CI,
std::unique_ptr<CompilerInstance> DCI);
@@ -175,10 +179,11 @@ class Interpreter {
llvm::Expected<Expr *> ExtractValueFromExpr(Expr *E);
llvm::Expected<llvm::orc::ExecutorAddr> CompileDtorCall(CXXRecordDecl
*CXXRD);
- CodeGenerator *getCodeGen() const;
- std::unique_ptr<llvm::Module> GenModule();
+ CodeGenerator *getCodeGen(IncrementalAction *Action = nullptr) const;
+ std::unique_ptr<llvm::Module> GenModule(IncrementalAction *Action = nullptr);
PartialTranslationUnit &RegisterPTU(TranslationUnitDecl *TU,
- std::unique_ptr<llvm::Module> M = {});
+ std::unique_ptr<llvm::Module> M = {},
+ IncrementalAction *Action = nullptr);
// A cache for the compiled destructors used to for de-allocation of managed
// clang::Values.
diff --git a/clang/lib/Interpreter/DeviceOffload.cpp
b/clang/lib/Interpreter/DeviceOffload.cpp
index 1999d63d1aa04..6977d7fa674ab 100644
--- a/clang/lib/Interpreter/DeviceOffload.cpp
+++ b/clang/lib/Interpreter/DeviceOffload.cpp
@@ -28,20 +28,21 @@ IncrementalCUDADeviceParser::IncrementalCUDADeviceParser(
std::unique_ptr<CompilerInstance> DeviceInstance,
CompilerInstance &HostInstance,
llvm::IntrusiveRefCntPtr<llvm::vfs::InMemoryFileSystem> FS,
- llvm::Error &Err, const std::list<PartialTranslationUnit> &PTUs)
+ llvm::Error &Err, std::list<PartialTranslationUnit> &PTUs)
: IncrementalParser(*DeviceInstance, Err), PTUs(PTUs), VFS(FS),
CodeGenOpts(HostInstance.getCodeGenOpts()),
- TargetOpts(HostInstance.getTargetOpts()) {
+ TargetOpts(DeviceInstance->getTargetOpts()) {
if (Err)
return;
- DeviceCI = std::move(DeviceInstance);
StringRef Arch = TargetOpts.CPU;
if (!Arch.starts_with("sm_") || Arch.substr(3).getAsInteger(10, SMVersion)) {
+ DeviceInstance.release();
Err = llvm::joinErrors(std::move(Err), llvm::make_error<llvm::StringError>(
"Invalid CUDA architecture",
llvm::inconvertibleErrorCode()));
return;
}
+ DeviceCI = std::move(DeviceInstance);
}
llvm::Expected<TranslationUnitDecl *>
@@ -50,25 +51,6 @@ IncrementalCUDADeviceParser::Parse(llvm::StringRef Input) {
if (!PTU)
return PTU.takeError();
- auto PTX = GeneratePTX();
- if (!PTX)
- return PTX.takeError();
-
- auto Err = GenerateFatbinary();
- if (Err)
- return std::move(Err);
-
- std::string FatbinFileName =
- "/incr_module_" + std::to_string(PTUs.size()) + ".fatbin";
- VFS->addFile(FatbinFileName, 0,
- llvm::MemoryBuffer::getMemBuffer(
- llvm::StringRef(FatbinContent.data(), FatbinContent.size()),
- "", false));
-
- CodeGenOpts.CudaGpuBinaryFileName = FatbinFileName;
-
- FatbinContent.clear();
-
return PTU;
}
@@ -78,9 +60,11 @@ llvm::Expected<llvm::StringRef>
IncrementalCUDADeviceParser::GeneratePTX() {
const llvm::Target *Target = llvm::TargetRegistry::lookupTarget(
PTU.TheModule->getTargetTriple(), Error);
- if (!Target)
+ if (!Target) {
return llvm::make_error<llvm::StringError>(std::move(Error),
std::error_code());
+ }
+
llvm::TargetOptions TO = llvm::TargetOptions();
llvm::TargetMachine *TargetMachine = Target->createTargetMachine(
PTU.TheModule->getTargetTriple(), TargetOpts.CPU, "", TO,
@@ -172,6 +156,19 @@ llvm::Error
IncrementalCUDADeviceParser::GenerateFatbinary() {
FatbinContent.append(PTXCode.begin(), PTXCode.end());
+ auto &PTU = PTUs.back();
+
+ std::string FatbinFileName = "/" + PTU.TheModule->getName().str() +
".fatbin";
+
+ VFS->addFile(FatbinFileName, 0,
+ llvm::MemoryBuffer::getMemBuffer(
+ llvm::StringRef(FatbinContent.data(), FatbinContent.size()),
+ "", false));
+
+ CodeGenOpts.CudaGpuBinaryFileName = FatbinFileName;
+
+ FatbinContent.clear();
+
return llvm::Error::success();
}
diff --git a/clang/lib/Interpreter/DeviceOffload.h
b/clang/lib/Interpreter/DeviceOffload.h
index b9a1acab004c3..23d89046c09e1 100644
--- a/clang/lib/Interpreter/DeviceOffload.h
+++ b/clang/lib/Interpreter/DeviceOffload.h
@@ -24,14 +24,14 @@ class CodeGenOptions;
class TargetOptions;
class IncrementalCUDADeviceParser : public IncrementalParser {
- const std::list<PartialTranslationUnit> &PTUs;
+ std::list<PartialTranslationUnit> &PTUs;
public:
IncrementalCUDADeviceParser(
std::unique_ptr<CompilerInstance> DeviceInstance,
CompilerInstance &HostInstance,
llvm::IntrusiveRefCntPtr<llvm::vfs::InMemoryFileSystem> VFS,
- llvm::Error &Err, const std::list<PartialTranslationUnit> &PTUs);
+ llvm::Error &Err, std::list<PartialTranslationUnit> &PTUs);
llvm::Expected<TranslationUnitDecl *> Parse(llvm::StringRef Input) override;
diff --git a/clang/lib/Interpreter/Interpreter.cpp
b/clang/lib/Interpreter/Interpreter.cpp
index f8c8d0a425659..e0865c6e9dfdf 100644
--- a/clang/lib/Interpreter/Interpreter.cpp
+++ b/clang/lib/Interpreter/Interpreter.cpp
@@ -451,13 +451,44 @@ const char *const Runtimes = R"(
)";
llvm::Expected<std::unique_ptr<Interpreter>>
-Interpreter::create(std::unique_ptr<CompilerInstance> CI) {
+Interpreter::create(std::unique_ptr<CompilerInstance> CI,
+ std::unique_ptr<CompilerInstance> DeviceCI) {
llvm::Error Err = llvm::Error::success();
auto Interp =
std::unique_ptr<Interpreter>(new Interpreter(std::move(CI), Err));
if (Err)
return std::move(Err);
+ CompilerInstance &HostCI = *(Interp->getCompilerInstance());
+
+ if (DeviceCI) {
+ Interp->DeviceAct = std::make_unique<IncrementalAction>(
+ *DeviceCI, *Interp->TSCtx->getContext(), Err, *Interp);
+
+ if (Err)
+ return std::move(Err);
+
+ DeviceCI->ExecuteAction(*Interp->DeviceAct);
+
+ // avoid writing fat binary to disk using an in-memory virtual file system
+ llvm::IntrusiveRefCntPtr<llvm::vfs::InMemoryFileSystem> IMVFS =
+ std::make_unique<llvm::vfs::InMemoryFileSystem>();
+ llvm::IntrusiveRefCntPtr<llvm::vfs::OverlayFileSystem> OverlayVFS =
+ std::make_unique<llvm::vfs::OverlayFileSystem>(
+ llvm::vfs::getRealFileSystem());
+ OverlayVFS->pushOverlay(IMVFS);
+ HostCI.createFileManager(OverlayVFS);
+
+ auto DeviceParser = std::make_unique<IncrementalCUDADeviceParser>(
+ std::move(DeviceCI), HostCI, IMVFS, Err,
+ Interp->PTUs);
+
+ if (Err)
+ return std::move(Err);
+
+ Interp->DeviceParser = std::move(DeviceParser);
+ }
+
// Add runtime code and set a marker to hide it from user code. Undo will not
// go through that.
auto PTU = Interp->Parse(Runtimes);
@@ -472,29 +503,7 @@ Interpreter::create(std::unique_ptr<CompilerInstance> CI) {
llvm::Expected<std::unique_ptr<Interpreter>>
Interpreter::createWithCUDA(std::unique_ptr<CompilerInstance> CI,
std::unique_ptr<CompilerInstance> DCI) {
- // avoid writing fat binary to disk using an in-memory virtual file system
- llvm::IntrusiveRefCntPtr<llvm::vfs::InMemoryFileSystem> IMVFS =
- std::make_unique<llvm::vfs::InMemoryFileSystem>();
- llvm::IntrusiveRefCntPtr<llvm::vfs::OverlayFileSystem> OverlayVFS =
- std::make_unique<llvm::vfs::OverlayFileSystem>(
- llvm::vfs::getRealFileSystem());
- OverlayVFS->pushOverlay(IMVFS);
- CI->createFileManager(OverlayVFS);
-
- auto Interp = Interpreter::create(std::move(CI));
- if (auto E = Interp.takeError())
- return std::move(E);
-
- llvm::Error Err = llvm::Error::success();
- auto DeviceParser = std::make_unique<IncrementalCUDADeviceParser>(
- std::move(DCI), *(*Interp)->getCompilerInstance(), IMVFS, Err,
- (*Interp)->PTUs);
- if (Err)
- return std::move(Err);
-
- (*Interp)->DeviceParser = std::move(DeviceParser);
-
- return Interp;
+ return Interpreter::create(std::move(CI), std::move(DCI));
}
const CompilerInstance *Interpreter::getCompilerInstance() const {
@@ -532,15 +541,16 @@ size_t Interpreter::getEffectivePTUSize() const {
PartialTranslationUnit &
Interpreter::RegisterPTU(TranslationUnitDecl *TU,
- std::unique_ptr<llvm::Module> M /*={}*/) {
+ std::unique_ptr<llvm::Module> M /*={}*/,
+ IncrementalAction *Action) {
PTUs.emplace_back(PartialTranslationUnit());
PartialTranslationUnit &LastPTU = PTUs.back();
LastPTU.TUPart = TU;
if (!M)
- M = GenModule();
+ M = GenModule(Action);
- assert((!getCodeGen() || M) && "Must have a llvm::Module at this point");
+ assert((!getCodeGen(Action) || M) && "Must have a llvm::Module at this
point");
LastPTU.TheModule = std::move(M);
LLVM_DEBUG(llvm::dbgs() << "compile-ptu " << PTUs.size() - 1
@@ -558,8 +568,21 @@ Interpreter::Parse(llvm::StringRef Code) {
// included in the host compilation
if (DeviceParser) {
llvm::Expected<TranslationUnitDecl *> DeviceTU = DeviceParser->Parse(Code);
- if (auto E = DeviceTU.takeError())
+ if (auto E = DeviceTU.takeError()) {
return std::move(E);
+ }
+
+ auto *CudaParser =
llvm::cast<IncrementalCUDADeviceParser>(DeviceParser.get());
+
+ PartialTranslationUnit &DevicePTU = RegisterPTU(*DeviceTU, nullptr,
DeviceAct.get());
+
+ llvm::Expected<llvm::StringRef> PTX = CudaParser->GeneratePTX();
+ if (!PTX)
+ return PTX.takeError();
+
+ llvm::Error Err = CudaParser->GenerateFatbinary();
+ if (Err)
+ return std::move(Err);
}
// Tell the interpreter sliently ignore unused expressions since value
@@ -736,9 +759,9 @@ llvm::Error Interpreter::LoadDynamicLibrary(const char
*name) {
return llvm::Error::success();
}
-std::unique_ptr<llvm::Module> Interpreter::GenModule() {
+std::unique_ptr<llvm::Module> Interpreter::GenModule(IncrementalAction
*Action) {
static unsigned ID = 0;
- if (CodeGenerator *CG = getCodeGen()) {
+ if (CodeGenerator *CG = getCodeGen(Action)) {
// Clang's CodeGen is designed to work with a single llvm::Module. In many
// cases for convenience various CodeGen parts have a reference to the
// llvm::Module (TheModule or Module) which does not change when a new
@@ -760,8 +783,10 @@ std::unique_ptr<llvm::Module> Interpreter::GenModule() {
return nullptr;
}
-CodeGenerator *Interpreter::getCodeGen() const {
- FrontendAction *WrappedAct = Act->getWrapped();
+CodeGenerator *Interpreter::getCodeGen(IncrementalAction *Action) const {
+ if (!Action)
+ Action = Act.get();
+ FrontendAction *WrappedAct = Action->getWrapped();
if (!WrappedAct->hasIRSupport())
return nullptr;
return static_cast<CodeGenAction *>(WrappedAct)->getCodeGenerator();
>From 9955a9971c83d7e9c5d138f593c1cb3fcbd68257 Mon Sep 17 00:00:00 2001
From: anutosh491 <andersonbhat...@gmail.com>
Date: Thu, 24 Apr 2025 17:32:23 +0530
Subject: [PATCH 2/3] Enable virtual file system to be used
---
clang/include/clang/Interpreter/Interpreter.h | 3 +-
clang/lib/Interpreter/DeviceOffload.cpp | 4 +-
clang/lib/Interpreter/Interpreter.cpp | 71 ++++++++++---------
3 files changed, 40 insertions(+), 38 deletions(-)
diff --git a/clang/include/clang/Interpreter/Interpreter.h
b/clang/include/clang/Interpreter/Interpreter.h
index 1b228e0917d02..59089cf639f96 100644
--- a/clang/include/clang/Interpreter/Interpreter.h
+++ b/clang/include/clang/Interpreter/Interpreter.h
@@ -132,8 +132,7 @@ class Interpreter {
public:
virtual ~Interpreter();
static llvm::Expected<std::unique_ptr<Interpreter>>
- create(std::unique_ptr<CompilerInstance> CI,
- std::unique_ptr<CompilerInstance> DeviceCI = nullptr);
+ create(std::unique_ptr<CompilerInstance> CI);
static llvm::Expected<std::unique_ptr<Interpreter>>
createWithCUDA(std::unique_ptr<CompilerInstance> CI,
std::unique_ptr<CompilerInstance> DCI);
diff --git a/clang/lib/Interpreter/DeviceOffload.cpp
b/clang/lib/Interpreter/DeviceOffload.cpp
index 6977d7fa674ab..d9b00787f038d 100644
--- a/clang/lib/Interpreter/DeviceOffload.cpp
+++ b/clang/lib/Interpreter/DeviceOffload.cpp
@@ -60,11 +60,9 @@ llvm::Expected<llvm::StringRef>
IncrementalCUDADeviceParser::GeneratePTX() {
const llvm::Target *Target = llvm::TargetRegistry::lookupTarget(
PTU.TheModule->getTargetTriple(), Error);
- if (!Target) {
+ if (!Target)
return llvm::make_error<llvm::StringError>(std::move(Error),
std::error_code());
- }
-
llvm::TargetOptions TO = llvm::TargetOptions();
llvm::TargetMachine *TargetMachine = Target->createTargetMachine(
PTU.TheModule->getTargetTriple(), TargetOpts.CPU, "", TO,
diff --git a/clang/lib/Interpreter/Interpreter.cpp
b/clang/lib/Interpreter/Interpreter.cpp
index e0865c6e9dfdf..a01dc93526635 100644
--- a/clang/lib/Interpreter/Interpreter.cpp
+++ b/clang/lib/Interpreter/Interpreter.cpp
@@ -451,44 +451,13 @@ const char *const Runtimes = R"(
)";
llvm::Expected<std::unique_ptr<Interpreter>>
-Interpreter::create(std::unique_ptr<CompilerInstance> CI,
- std::unique_ptr<CompilerInstance> DeviceCI) {
+Interpreter::create(std::unique_ptr<CompilerInstance> CI) {
llvm::Error Err = llvm::Error::success();
auto Interp =
std::unique_ptr<Interpreter>(new Interpreter(std::move(CI), Err));
if (Err)
return std::move(Err);
- CompilerInstance &HostCI = *(Interp->getCompilerInstance());
-
- if (DeviceCI) {
- Interp->DeviceAct = std::make_unique<IncrementalAction>(
- *DeviceCI, *Interp->TSCtx->getContext(), Err, *Interp);
-
- if (Err)
- return std::move(Err);
-
- DeviceCI->ExecuteAction(*Interp->DeviceAct);
-
- // avoid writing fat binary to disk using an in-memory virtual file system
- llvm::IntrusiveRefCntPtr<llvm::vfs::InMemoryFileSystem> IMVFS =
- std::make_unique<llvm::vfs::InMemoryFileSystem>();
- llvm::IntrusiveRefCntPtr<llvm::vfs::OverlayFileSystem> OverlayVFS =
- std::make_unique<llvm::vfs::OverlayFileSystem>(
- llvm::vfs::getRealFileSystem());
- OverlayVFS->pushOverlay(IMVFS);
- HostCI.createFileManager(OverlayVFS);
-
- auto DeviceParser = std::make_unique<IncrementalCUDADeviceParser>(
- std::move(DeviceCI), HostCI, IMVFS, Err,
- Interp->PTUs);
-
- if (Err)
- return std::move(Err);
-
- Interp->DeviceParser = std::move(DeviceParser);
- }
-
// Add runtime code and set a marker to hide it from user code. Undo will not
// go through that.
auto PTU = Interp->Parse(Runtimes);
@@ -503,7 +472,43 @@ Interpreter::create(std::unique_ptr<CompilerInstance> CI,
llvm::Expected<std::unique_ptr<Interpreter>>
Interpreter::createWithCUDA(std::unique_ptr<CompilerInstance> CI,
std::unique_ptr<CompilerInstance> DCI) {
- return Interpreter::create(std::move(CI), std::move(DCI));
+ // avoid writing fat binary to disk using an in-memory virtual file system
+ llvm::IntrusiveRefCntPtr<llvm::vfs::InMemoryFileSystem> IMVFS =
+ std::make_unique<llvm::vfs::InMemoryFileSystem>();
+ llvm::IntrusiveRefCntPtr<llvm::vfs::OverlayFileSystem> OverlayVFS =
+ std::make_unique<llvm::vfs::OverlayFileSystem>(
+ llvm::vfs::getRealFileSystem());
+ OverlayVFS->pushOverlay(IMVFS);
+ CI->createFileManager(OverlayVFS);
+
+ llvm::Expected<std::unique_ptr<Interpreter>> InterpOrErr =
Interpreter::create(std::move(CI));
+ if (!InterpOrErr)
+ return InterpOrErr;
+
+ std::unique_ptr<Interpreter> Interp = std::move(*InterpOrErr);
+
+ llvm::Error Err = llvm::Error::success();
+ llvm::LLVMContext &LLVMCtx = *Interp->TSCtx->getContext();
+
+ auto DeviceAct = std::make_unique<IncrementalAction>(
+ *DCI, LLVMCtx, Err, *Interp);
+
+ if (Err)
+ return std::move(Err);
+
+ Interp->DeviceAct = std::move(DeviceAct);
+
+ DCI->ExecuteAction(*Interp->DeviceAct);
+
+ auto DeviceParser = std::make_unique<IncrementalCUDADeviceParser>(
+ std::move(DCI), *Interp->getCompilerInstance(), IMVFS, Err,
+ Interp->PTUs);
+
+ if (Err)
+ return std::move(Err);
+
+ Interp->DeviceParser = std::move(DeviceParser);
+ return std::move(Interp);
}
const CompilerInstance *Interpreter::getCompilerInstance() const {
>From 77c773e8caa2dd723406b1feda18259bf47e7105 Mon Sep 17 00:00:00 2001
From: anutosh491 <andersonbhat...@gmail.com>
Date: Thu, 24 Apr 2025 17:40:41 +0530
Subject: [PATCH 3/3] remove unused variable
---
clang/lib/Interpreter/Interpreter.cpp | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/clang/lib/Interpreter/Interpreter.cpp
b/clang/lib/Interpreter/Interpreter.cpp
index a01dc93526635..4a541188c0557 100644
--- a/clang/lib/Interpreter/Interpreter.cpp
+++ b/clang/lib/Interpreter/Interpreter.cpp
@@ -579,7 +579,7 @@ Interpreter::Parse(llvm::StringRef Code) {
auto *CudaParser =
llvm::cast<IncrementalCUDADeviceParser>(DeviceParser.get());
- PartialTranslationUnit &DevicePTU = RegisterPTU(*DeviceTU, nullptr,
DeviceAct.get());
+ RegisterPTU(*DeviceTU, nullptr, DeviceAct.get());
llvm::Expected<llvm::StringRef> PTX = CudaParser->GeneratePTX();
if (!PTX)
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
