[C++20] [Modules] Introduce -fgen-reduced-bmi

Author: ChuanqiXu9

clang/include/clang/CodeGen/CodeGenAction.h

@@ -57,6 +57,8 @@ class CodeGenAction : public ASTFrontendAction {
   bool loadLinkModules(CompilerInstance &CI);
 
 protected:
+  bool BeginSourceFileAction(CompilerInstance &CI) override;
+
   /// Create a new code generation action.  If the optional \p _VMContext
   /// parameter is supplied, the action uses it without taking ownership,
   /// otherwise it creates a fresh LLVM context and takes ownership.

clang/include/clang/Driver/Options.td

@@ -3018,6 +3018,7 @@ defm prebuilt_implicit_modules : BoolFOption<"prebuilt-implicit-modules",
 
 def fmodule_output_EQ : Joined<["-"], "fmodule-output=">,
   Flags<[NoXarchOption]>, Visibility<[ClangOption, CC1Option]>,
+  MarshallingInfoString<FrontendOpts<"ModuleOutputPath">>,
   HelpText<"Save intermediate module file results when compiling a standard C++ module unit.">;
 def fmodule_output : Flag<["-"], "fmodule-output">, Flags<[NoXarchOption]>,
   Visibility<[ClangOption, CC1Option]>,
@@ -3031,6 +3032,11 @@ defm skip_odr_check_in_gmf : BoolOption<"f", "skip-odr-check-in-gmf",
           "Perform ODR checks for decls in the global module fragment.">>,
   Group<f_Group>;
 
+def gen_reduced_bmi : Flag<["-"], "fgen-reduced-bmi">,
+  Group<i_Group>, Visibility<[ClangOption, CC1Option]>,
+  HelpText<"Generate the reduced BMI">,
+  MarshallingInfoFlag<FrontendOpts<"GenReducedBMI">>;
+
 def fmodules_prune_interval : Joined<["-"], "fmodules-prune-interval=">, Group<i_Group>,
   Visibility<[ClangOption, CC1Option]>, MetaVarName<"<seconds>">,
   HelpText<"Specify the interval (in seconds) between attempts to prune the module cache">,

clang/include/clang/Frontend/FrontendOptions.h

@@ -387,6 +387,10 @@ class FrontendOptions {
   LLVM_PREFERRED_TYPE(bool)
   unsigned ModulesShareFileManager : 1;
 
+  /// Whether to generate reduced BMI for C++20 named modules.
+  LLVM_PREFERRED_TYPE(bool)
+  unsigned GenReducedBMI : 1;
+
   CodeCompleteOptions CodeCompleteOpts;
 
   /// Specifies the output format of the AST.
@@ -553,6 +557,9 @@ class FrontendOptions {
   /// Path which stores the output files for -ftime-trace
   std::string TimeTracePath;
 
+  /// Output Path for module output file.
+  std::string ModuleOutputPath;
+
 public:
   FrontendOptions()
       : DisableFree(false), RelocatablePCH(false), ShowHelp(false),
@@ -565,7 +572,7 @@ class FrontendOptions {
         BuildingImplicitModuleUsesLock(true), ModulesEmbedAllFiles(false),
         IncludeTimestamps(true), UseTemporary(true),
         AllowPCMWithCompilerErrors(false), ModulesShareFileManager(true),
-        TimeTraceGranularity(500) {}
+        GenReducedBMI(false), TimeTraceGranularity(500) {}
 
   /// getInputKindForExtension - Return the appropriate input kind for a file
   /// extension. For example, "c" would return Language::C.

clang/include/clang/Serialization/ASTWriter.h

@@ -846,7 +846,7 @@ class ASTWriter : public ASTDeserializationListener,
 /// AST and semantic-analysis consumer that generates a
 /// precompiled header from the parsed source code.
 class PCHGenerator : public SemaConsumer {
-  const Preprocessor &PP;
+  Preprocessor &PP;
   std::string OutputFile;
   std::string isysroot;
   Sema *SemaPtr;
@@ -867,11 +867,12 @@ class PCHGenerator : public SemaConsumer {
   DiagnosticsEngine &getDiagnostics() const {
     return SemaPtr->getDiagnostics();
   }
+  Preprocessor &getPreprocessor() { return PP; }
 
   virtual Module *getEmittingModule(ASTContext &Ctx);
 
 public:
-  PCHGenerator(const Preprocessor &PP, InMemoryModuleCache &ModuleCache,
+  PCHGenerator(Preprocessor &PP, InMemoryModuleCache &ModuleCache,
                StringRef OutputFile, StringRef isysroot,
                std::shared_ptr<PCHBuffer> Buffer,
                ArrayRef<std::shared_ptr<ModuleFileExtension>> Extensions,
@@ -893,7 +894,7 @@ class ReducedBMIGenerator : public PCHGenerator {
   virtual Module *getEmittingModule(ASTContext &Ctx) override;
 
 public:
-  ReducedBMIGenerator(const Preprocessor &PP, InMemoryModuleCache &ModuleCache,
+  ReducedBMIGenerator(Preprocessor &PP, InMemoryModuleCache &ModuleCache,
                       StringRef OutputFile);
 
   void HandleTranslationUnit(ASTContext &Ctx) override;

clang/lib/CodeGen/CodeGenAction.cpp

@@ -25,8 +25,11 @@
 #include "clang/CodeGen/ModuleBuilder.h"
 #include "clang/Driver/DriverDiagnostic.h"
 #include "clang/Frontend/CompilerInstance.h"
+#include "clang/Frontend/FrontendActions.h"
 #include "clang/Frontend/FrontendDiagnostic.h"
+#include "clang/Frontend/MultiplexConsumer.h"
 #include "clang/Lex/Preprocessor.h"
+#include "clang/Serialization/ASTWriter.h"
 #include "llvm/ADT/Hashing.h"
 #include "llvm/Bitcode/BitcodeReader.h"
 #include "llvm/CodeGen/MachineOptimizationRemarkEmitter.h"
@@ -1003,6 +1006,12 @@ CodeGenerator *CodeGenAction::getCodeGenerator() const {
   return BEConsumer->getCodeGenerator();
 }
 
+bool CodeGenAction::BeginSourceFileAction(CompilerInstance &CI) {
+  if (CI.getFrontendOpts().GenReducedBMI)
+    CI.getLangOpts().setCompilingModule(LangOptions::CMK_ModuleInterface);
+  return true;
+}
+
 static std::unique_ptr<raw_pwrite_stream>
 GetOutputStream(CompilerInstance &CI, StringRef InFile, BackendAction Action) {
   switch (Action) {
@@ -1061,6 +1070,16 @@ CodeGenAction::CreateASTConsumer(CompilerInstance &CI, StringRef InFile) {
     CI.getPreprocessor().addPPCallbacks(std::move(Callbacks));
   }
 
+  if (CI.getFrontendOpts().GenReducedBMI &&
+      !CI.getFrontendOpts().ModuleOutputPath.empty()) {
+    std::vector<std::unique_ptr<ASTConsumer>> Consumers{2};
+    Consumers[0] = std::move(Result);
+    Consumers[1] = std::make_unique<ReducedBMIGenerator>(
+        CI.getPreprocessor(), CI.getModuleCache(),
+        CI.getFrontendOpts().ModuleOutputPath);
+    return std::make_unique<MultiplexConsumer>(std::move(Consumers));
+  }
+
   return std::move(Result);
 }
 

clang/lib/Driver/Driver.cpp

@@ -4737,6 +4737,14 @@ Action *Driver::ConstructPhaseAction(
     if (Args.hasArg(options::OPT_extract_api))
       return C.MakeAction<ExtractAPIJobAction>(Input, types::TY_API_INFO);
 
+    // With '-fgen-reduced-bmi', we don't want to run the precompile phase
+    // unless the user specified '--precompile'. In the case the '--precompile'
+    // flag is enabled, we will try to emit the reduced BMI as a by product
+    // in GenerateModuleInterfaceAction.
+    if (Args.hasArg(options::OPT_gen_reduced_bmi) &&
+        !Args.getLastArg(options::OPT__precompile))
+      return Input;
+
     types::ID OutputTy = getPrecompiledType(Input->getType());
     assert(OutputTy != types::TY_INVALID &&
            "Cannot precompile this input type!");
@@ -5802,19 +5810,8 @@ static const char *GetModuleOutputPath(Compilation &C, const JobAction &JA,
          (C.getArgs().hasArg(options::OPT_fmodule_output) ||
           C.getArgs().hasArg(options::OPT_fmodule_output_EQ)));
 
-  if (Arg *ModuleOutputEQ =
-          C.getArgs().getLastArg(options::OPT_fmodule_output_EQ))
-    return C.addResultFile(ModuleOutputEQ->getValue(), &JA);
-
-  SmallString<64> OutputPath;
-  Arg *FinalOutput = C.getArgs().getLastArg(options::OPT_o);
-  if (FinalOutput && C.getArgs().hasArg(options::OPT_c))
-    OutputPath = FinalOutput->getValue();
-  else
-    OutputPath = BaseInput;
-
-  const char *Extension = types::getTypeTempSuffix(JA.getType());
-  llvm::sys::path::replace_extension(OutputPath, Extension);
+  SmallString<256> OutputPath =
+      tools::getCXX20NamedModuleOutputPath(C.getArgs(), BaseInput);
   return C.addResultFile(C.getArgs().MakeArgString(OutputPath.c_str()), &JA);
 }
 
@@ -5901,8 +5898,10 @@ const char *Driver::GetNamedOutputPath(Compilation &C, const JobAction &JA,
   // If we're emitting a module output with the specified option
   // `-fmodule-output`.
   if (!AtTopLevel && isa<PrecompileJobAction>(JA) &&
-      JA.getType() == types::TY_ModuleFile && SpecifiedModuleOutput)
+      JA.getType() == types::TY_ModuleFile && SpecifiedModuleOutput) {
+    assert(!C.getArgs().hasArg(options::OPT_gen_reduced_bmi));
     return GetModuleOutputPath(C, JA, BaseInput);
+  }
 
   // Output to a temporary file?
   if ((!AtTopLevel && !isSaveTempsEnabled() &&

clang/lib/Driver/ToolChains/Clang.cpp

@@ -3777,6 +3777,24 @@ bool Driver::getDefaultModuleCachePath(SmallVectorImpl<char> &Result) {
   return false;
 }
 
+llvm::SmallString<256>
+clang::driver::tools::getCXX20NamedModuleOutputPath(const ArgList &Args,
+                                                    const char *BaseInput) {
+  if (Arg *ModuleOutputEQ = Args.getLastArg(options::OPT_fmodule_output_EQ))
+    return StringRef(ModuleOutputEQ->getValue());
+
+  SmallString<256> OutputPath;
+  if (Arg *FinalOutput = Args.getLastArg(options::OPT_o);
+      FinalOutput && Args.hasArg(options::OPT_c))
+    OutputPath = FinalOutput->getValue();
+  else
+    OutputPath = BaseInput;
+
+  const char *Extension = types::getTypeTempSuffix(types::TY_ModuleFile);
+  llvm::sys::path::replace_extension(OutputPath, Extension);
+  return OutputPath;
+}
+
 static bool RenderModulesOptions(Compilation &C, const Driver &D,
                                  const ArgList &Args, const InputInfo &Input,
                                  const InputInfo &Output, bool HaveStd20,
@@ -3965,9 +3983,26 @@ static bool RenderModulesOptions(Compilation &C, const Driver &D,
   // module fragment.
   CmdArgs.push_back("-fskip-odr-check-in-gmf");
 
-  // Claim `-fmodule-output` and `-fmodule-output=` to avoid unused warnings.
-  Args.ClaimAllArgs(options::OPT_fmodule_output);
-  Args.ClaimAllArgs(options::OPT_fmodule_output_EQ);
+  // Noop if we see '-fgen-reduced-bmi' with other translation units than module
+  // units. This is more user friendly to allow end uers to enable this feature
+  // without asking for help from build systems.
+  if (Args.hasArg(options::OPT_gen_reduced_bmi) &&
+      (Input.getType() == driver::types::TY_CXXModule ||
+       Input.getType() == driver::types::TY_PP_CXXModule)) {
+    CmdArgs.push_back("-fgen-reduced-bmi");
+
+    if (Args.hasArg(options::OPT_fmodule_output_EQ))
+      Args.AddLastArg(CmdArgs, options::OPT_fmodule_output_EQ);
+    else
+      CmdArgs.push_back(Args.MakeArgString(
+          "-fmodule-output=" +
+          getCXX20NamedModuleOutputPath(Args, Input.getBaseInput())));
+  } else {
+    // To avoid unused warnings.
+    Args.ClaimAllArgs(options::OPT_gen_reduced_bmi);
+    Args.ClaimAllArgs(options::OPT_fmodule_output);
+    Args.ClaimAllArgs(options::OPT_fmodule_output_EQ);
+  }
 
   return HaveModules;
 }

clang/lib/Driver/ToolChains/Clang.h

@@ -193,6 +193,20 @@ DwarfFissionKind getDebugFissionKind(const Driver &D,
                                      const llvm::opt::ArgList &Args,
                                      llvm::opt::Arg *&Arg);
 
+// Calculate the output path of the module file when compiling a module unit
+// with the `-fmodule-output` option or `-fmodule-output=` option specified.
+// The behavior is:
+// - If `-fmodule-output=` is specfied, then the module file is
+//   writing to the value.
+// - Otherwise if the output object file of the module unit is specified, the
+// output path
+//   of the module file should be the same with the output object file except
+//   the corresponding suffix. This requires both `-o` and `-c` are specified.
+// - Otherwise, the output path of the module file will be the same with the
+//   input with the corresponding suffix.
+llvm::SmallString<256>
+getCXX20NamedModuleOutputPath(const llvm::opt::ArgList &Args,
+                              const char *BaseInput);
 } // end namespace tools
 
 } // end namespace driver

clang/lib/Frontend/FrontendActions.cpp

@@ -281,6 +281,13 @@ GenerateModuleInterfaceAction::CreateASTConsumer(CompilerInstance &CI,
   if (Consumers.empty())
     return nullptr;
 
+  if (CI.getFrontendOpts().GenReducedBMI &&
+      !CI.getFrontendOpts().ModuleOutputPath.empty()) {
+    Consumers.push_back(std::make_unique<ReducedBMIGenerator>(
+        CI.getPreprocessor(), CI.getModuleCache(),
+        CI.getFrontendOpts().ModuleOutputPath));
+  }
+
   return std::make_unique<MultiplexConsumer>(std::move(Consumers));
 }
 

clang/lib/Frontend/PrecompiledPreamble.cpp

@@ -290,8 +290,7 @@ class PrecompilePreambleAction : public ASTFrontendAction {
 
 class PrecompilePreambleConsumer : public PCHGenerator {
 public:
-  PrecompilePreambleConsumer(PrecompilePreambleAction &Action,
-                             const Preprocessor &PP,
+  PrecompilePreambleConsumer(PrecompilePreambleAction &Action, Preprocessor &PP,
                              InMemoryModuleCache &ModuleCache,
                              StringRef isysroot,
                              std::shared_ptr<PCHBuffer> Buffer)

clang/lib/Serialization/GeneratePCH.cpp

@@ -14,6 +14,7 @@
 #include "clang/AST/ASTContext.h"
 #include "clang/Frontend/FrontendDiagnostic.h"
 #include "clang/Lex/HeaderSearch.h"
+#include "clang/Lex/HeaderSearchOptions.h"
 #include "clang/Lex/Preprocessor.h"
 #include "clang/Sema/SemaConsumer.h"
 #include "clang/Serialization/ASTReader.h"
@@ -23,8 +24,8 @@
 using namespace clang;
 
 PCHGenerator::PCHGenerator(
-    const Preprocessor &PP, InMemoryModuleCache &ModuleCache,
-    StringRef OutputFile, StringRef isysroot, std::shared_ptr<PCHBuffer> Buffer,
+    Preprocessor &PP, InMemoryModuleCache &ModuleCache, StringRef OutputFile,
+    StringRef isysroot, std::shared_ptr<PCHBuffer> Buffer,
     ArrayRef<std::shared_ptr<ModuleFileExtension>> Extensions,
     bool AllowASTWithErrors, bool IncludeTimestamps,
     bool BuildingImplicitModule, bool ShouldCacheASTInMemory,
@@ -88,7 +89,7 @@ ASTDeserializationListener *PCHGenerator::GetASTDeserializationListener() {
   return &Writer;
 }
 
-ReducedBMIGenerator::ReducedBMIGenerator(const Preprocessor &PP,
+ReducedBMIGenerator::ReducedBMIGenerator(Preprocessor &PP,
                                          InMemoryModuleCache &ModuleCache,
                                          StringRef OutputFile)
     : PCHGenerator(
@@ -101,12 +102,26 @@ ReducedBMIGenerator::ReducedBMIGenerator(const Preprocessor &PP,
 
 Module *ReducedBMIGenerator::getEmittingModule(ASTContext &Ctx) {
   Module *M = Ctx.getCurrentNamedModule();
-  assert(M->isNamedModuleUnit() &&
+  assert(M && M->isNamedModuleUnit() &&
          "ReducedBMIGenerator should only be used with C++20 Named modules.");
   return M;
 }
 
 void ReducedBMIGenerator::HandleTranslationUnit(ASTContext &Ctx) {
+  // FIMXE: We'd better to wrap such options to a new class ASTWriterOptions.
+  getPreprocessor()
+      .getHeaderSearchInfo()
+      .getHeaderSearchOpts()
+      .ModulesSkipDiagnosticOptions = true;
+  getPreprocessor()
+      .getHeaderSearchInfo()
+      .getHeaderSearchOpts()
+      .ModulesSkipHeaderSearchPaths = true;
+  getPreprocessor()
+      .getHeaderSearchInfo()
+      .getHeaderSearchOpts()
+      .ModulesSkipPragmaDiagnosticMappings = true;
+
   PCHGenerator::HandleTranslationUnit(Ctx);
 
   if (!isComplete())

clang/test/Driver/module-fgen-reduced-bmi.cppm

@@ -0,0 +1,53 @@
+// It is annoying to handle different slash direction
+// in Windows and Linux. So we disable the test on Windows
+// here.
+// REQUIRES: !system-windows
+// On AIX, the default output for `-c` may be `.s` instead of `.o`,
+// which makes the test fail. So disable the test on AIX.
+// REQUIRES: !system-aix
+//
+// RUN: rm -rf %t
+// RUN: mkdir %t
+// RUN: split-file %s %t
+//
+// RUN: %clang -std=c++20 %t/Hello.cppm -fmodule-output=%t/Hello.pcm \
+// RUN:     -fgen-reduced-bmi -c -o %t/Hello.o -### 2>&1 | FileCheck %t/Hello.cppm
+//
+// RUN: %clang -std=c++20 %t/Hello.cppm \
+// RUN:     -fgen-reduced-bmi -c -o %t/Hello.o -### 2>&1 | \
+// RUN:         FileCheck %t/Hello.cppm --check-prefix=CHECK-UNSPECIFIED
+//
+// RUN: %clang -std=c++20 %t/Hello.cppm \
+// RUN:     -fgen-reduced-bmi -c -### 2>&1 | \
+// RUN:         FileCheck %t/Hello.cppm --check-prefix=CHECK-NO-O
+//
+// RUN: %clang -std=c++20 %t/Hello.cppm \
+// RUN:     -fgen-reduced-bmi -c -o %t/AnotherName.o -### 2>&1 | \
+// RUN:         FileCheck %t/Hello.cppm --check-prefix=CHECK-ANOTHER-NAME
+//
+// RUN: %clang -std=c++20 %t/Hello.cppm --precompile -fgen-reduced-bmi \
+// RUN:     -o %t/Hello.full.pcm -### 2>&1 | FileCheck %t/Hello.cppm \
+// RUN:     --check-prefix=CHECK-EMIT-MODULE-INTERFACE
+//
+// RUN: %clang -std=c++20 %t/Hello.cc -fgen-reduced-bmi -Wall -Werror \
+// RUN:     -c -o %t/Hello.o -### 2>&1 | FileCheck %t/Hello.cc
+
+//--- Hello.cppm
+export module Hello;
+
+// Test that we won't generate the emit-module-interface as 2 phase compilation model.
+// CHECK-NOT: -emit-module-interface
+// CHECK: "-fgen-reduced-bmi"
+
+// CHECK-UNSPECIFIED: -fmodule-output={{.*}}/Hello.pcm
+
+// CHECK-NO-O: -fmodule-output={{.*}}/Hello.pcm
+// CHECK-ANOTHER-NAME: -fmodule-output={{.*}}/AnotherName.pcm
+
+// With `-emit-module-interface` specified, we should still see the `-emit-module-interface`
+// flag.
+// CHECK-EMIT-MODULE-INTERFACE: -emit-module-interface
+
+//--- Hello.cc
+
+// CHECK-NOT: "-fgen-reduced-bmi"