[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

@@ -3017,6 +3017,7 @@ 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]>,
@@ -3030,6 +3031,11 @@ 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 modules_reduced_bmi : Flag<["-"], "fexperimental-modules-reduced-bmi">,
+  Group<f_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/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::make_unique<ReducedBMIGenerator>(
+        CI.getPreprocessor(), CI.getModuleCache(),
+        CI.getFrontendOpts().ModuleOutputPath);
+    Consumers[1] = std::move(Result);
+    return std::make_unique<MultiplexConsumer>(std::move(Consumers));
+  }
+
   return std::move(Result);
 }
 

clang/lib/Driver/Driver.cpp

@@ -4749,6 +4749,14 @@ Action *Driver::ConstructPhaseAction(
     if (Args.hasArg(options::OPT_extract_api))
       return C.MakeAction<ExtractAPIJobAction>(Input, types::TY_API_INFO);
 
+    // With 'fexperimental-modules-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_modules_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!");
@@ -5903,8 +5911,10 @@ 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_modules_reduced_bmi));
     return GetModuleOutputPath(C, JA, BaseInput);
+  }
 
   // Output to a temporary file?
   if ((!AtTopLevel && !isSaveTempsEnabled() &&

clang/lib/Driver/ToolChains/Clang.cpp

@@ -4045,6 +4045,24 @@ static bool RenderModulesOptions(Compilation &C, const Driver &D,
   // module fragment.
   CmdArgs.push_back("-fskip-odr-check-in-gmf");
 
+  if (Args.hasArg(options::OPT_modules_reduced_bmi) &&
+      (Input.getType() == driver::types::TY_CXXModule ||
+       Input.getType() == driver::types::TY_PP_CXXModule)) {
+    CmdArgs.push_back("-fexperimental-modules-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())));
+  }
+
+  // Noop if we see '-fexperimental-modules-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.
+  Args.ClaimAllArgs(options::OPT_modules_reduced_bmi);
+
   // We need to include the case the input file is a module file here.
   // Since the default compilation model for C++ module interface unit will
   // create temporary module file and compile the temporary module file

clang/lib/Frontend/FrontendActions.cpp

@@ -284,6 +284,13 @@ 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/test/Driver/module-fgen-reduced-bmi.cppm

@@ -0,0 +1,51 @@
+// 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.
+// UNSUPPORTED: system-aix
+//
+// RUN: rm -rf %t && split-file %s %t && cd %t
+//
+// RUN: %clang -std=c++20 Hello.cppm -fmodule-output=Hello.pcm \
+// RUN:     -fexperimental-modules-reduced-bmi -c -o Hello.o -### 2>&1 | FileCheck Hello.cppm
+//
+// RUN: %clang -std=c++20 Hello.cppm \
+// RUN:     -fexperimental-modules-reduced-bmi -c -o Hello.o -### 2>&1 | \
+// RUN:         FileCheck Hello.cppm --check-prefix=CHECK-UNSPECIFIED
+//
+// RUN: %clang -std=c++20 Hello.cppm \
+// RUN:     -fexperimental-modules-reduced-bmi -c -### 2>&1 | \
+// RUN:         FileCheck Hello.cppm --check-prefix=CHECK-NO-O
+//
+// RUN: %clang -std=c++20 Hello.cppm \
+// RUN:     -fexperimental-modules-reduced-bmi -c -o AnotherName.o -### 2>&1 | \
+// RUN:         FileCheck Hello.cppm --check-prefix=CHECK-ANOTHER-NAME
+//
+// RUN: %clang -std=c++20 Hello.cppm --precompile -fexperimental-modules-reduced-bmi \
+// RUN:     -o Hello.full.pcm -### 2>&1 | FileCheck Hello.cppm \
+// RUN:     --check-prefix=CHECK-EMIT-MODULE-INTERFACE
+//
+// RUN: %clang -std=c++20 Hello.cc -fexperimental-modules-reduced-bmi -Wall -Werror \
+// RUN:     -c -o Hello.o -### 2>&1 | FileCheck 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: "-fexperimental-modules-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: "-fexperimental-modules-reduced-bmi"

clang/test/Modules/modules-reduced-bmi.cppm

@@ -0,0 +1,36 @@
+// RUN: rm -rf %t
+// RUN: mkdir -p %t
+// RUN: split-file %s %t
+//
+// RUN: %clang_cc1 -std=c++20 %t/a.cppm -emit-reduced-module-interface -o %t/a.reduced.pcm
+// RUN: %clang_cc1 -std=c++20 %t/a.cppm -fexperimental-modules-reduced-bmi -fmodule-output=%t/a.pcm \
+// RUN:     -S -emit-llvm -o %t/a.ll
+//
+// Test that the generated BMI from `-fexperimental-modules-reduced-bmi -fmodule-output=` is same with
+// `-emit-reduced-module-interface`.
+// RUN: diff %t/a.reduced.pcm %t/a.pcm
+//
+// Test that we can consume the produced BMI correctly.
+// RUN: %clang_cc1 -std=c++20 %t/b.cppm -fmodule-file=a=%t/a.pcm -fsyntax-only -verify
+//
+// RUN: rm -f %t/a.pcm
+// RUN: %clang_cc1 -std=c++20 %t/a.cppm -fexperimental-modules-reduced-bmi -fmodule-output=%t/a.pcm \
+// RUN:     -emit-module-interface -o %t/a.full.pcm
+// RUN: diff %t/a.reduced.pcm %t/a.pcm
+// RUN: not diff %t/a.pcm %t/a.full.pcm
+//
+// RUN: %clang_cc1 -std=c++20 %t/b.cppm -fmodule-file=a=%t/a.pcm -fsyntax-only -verify
+// RUN: %clang_cc1 -std=c++20 %t/b.cppm -fmodule-file=a=%t/a.full.pcm -fsyntax-only -verify
+
+//--- a.cppm
+export module a;
+export int a() {
+    return 43;
+}
+
+//--- b.cppm
+// Test that we can consume the produced BMI correctly as a smocking test.
+// expected-no-diagnostics
+export module b;
+import a;
+export int b() { return a(); }