Introduce -fembed-bitcode driver option

[steven_wu]

Summary

This is the clang driver part of the change to embedded bitcode. This
includes:

1. -fembed-bitcode option which breaks down the compilation into two

stages. The first stage emits optimized bitcode and the second stage
compiles bitcode into object file.

2. -fembed-bitcode-marker option which doesn't really break down to

two stages to speedup the compilation flow.

3. pass the correct linker flag to darwin linker if tool chains supports

embedded bitcode.

[thakis]

I don't understand the changes to Driver.cpp lines 1760 and down. The rest looks good to me.

lib/Driver/Driver.cpp

1763

Why is this change needed? If its needed, is the preceding comment out of date now?

[steven_wu]

Make selectToolForJob logic better and less disruptive for -fembed-bitcode.

When -fembed-bitcode is used, the same old logic is used to decide if
CompileJob and BackendJob can be combined with an extra requirement for
-fembed-bitcode. The old logic was there to work around the Fortran input
which invokes gcc to compile. -fembed-bitcode requires CompileJob and BackendJob
to be the same (and can emit IR) so it can be splited between these two stages.

[thakis]

Thanks, that's much clearer :-)

[steven_wu]

Thanks for reviewing the patch. There is a problem in the test case (it only works on darwin). It is fixed in r262286.

lib/Driver/Driver.cpp

1763

The preceding comment is still true but this condition makes BackendInputs->begin() not always CompileJobAction. I have a cleaner way to handle this. I am rebasing and updating the patch.

include/clang/Basic/DiagnosticDriverKinds.td

 def err_drv_optimization_remark_pattern : Error<
   "%0 in '%1'">;
 def err_drv_no_neon_modifier : Error<"[no]neon is not accepted as modifier, please use [no]simd instead">;
 def err_drv_invalid_omp_target : Error<"OpenMP target is invalid: '%0'">;
 def err_drv_omp_host_ir_file_not_found : Error<
   "The provided host compiler IR file '%0' is required to generate code for OpenMP target regions but cannot be found.">;
 def err_drv_omp_host_target_not_supported : Error<
   "The target '%0' is not a supported OpenMP host target.">;
-  
+def err_drv_bitcode_unsupported_on_toolchain : Error<
+  "-fembed-bitcode is not supported on versions of iOS prior to 6.0">;
 def warn_O4_is_O3 : Warning<"-O4 is equivalent to -O3">, InGroup<Deprecated>;
 def warn_drv_lto_libpath : Warning<"libLTO.dylib relative to clang installed dir not found; using 'ld' default search path instead">,
   InGroup<LibLTO>;
 def warn_drv_optimization_value : Warning<"optimization level '%0' is not supported; using '%1%2' instead">,
   InGroup<InvalidCommandLineArgument>;
 def warn_ignored_gcc_optimization : Warning<"optimization flag '%0' is not supported">,
   InGroup<IgnoredOptimizationArgument>;
 def warn_c_kext : Warning<

include/clang/Driver/Driver.h

class Driver {
   } Mode;
 
   enum SaveTempsMode {
     SaveTempsNone,
     SaveTempsCwd,
     SaveTempsObj
   } SaveTemps;
 
+  enum BitcodeEmbedMode {
+    EmbedNone,
+    EmbedMarker,
+    EmbedBitcode
+  } BitcodeEmbed;
+
   /// LTO mode selected via -f(no-)?lto(=.*)? options.
   LTOKind LTOMode;
 
 public:
   // Diag - Forwarding function for diagnostics.
   DiagnosticBuilder Diag(unsigned DiagID) const {
     return Diags.Report(DiagID);
   }
public:
   }
   void setInstalledDir(StringRef Value) {
     InstalledDir = Value;
   }
 
   bool isSaveTempsEnabled() const { return SaveTemps != SaveTempsNone; }
   bool isSaveTempsObj() const { return SaveTemps == SaveTempsObj; }
 
+  bool embedBitcodeEnabled() const { return BitcodeEmbed == EmbedBitcode; }
+  bool embedBitcodeMarkerOnly() const { return BitcodeEmbed == EmbedMarker; }
+
   /// @}
   /// @name Primary Functionality
   /// @{
 
   /// BuildCompilation - Construct a compilation object for a command
   /// line argument vector.
   ///
   /// \return A compilation, or 0 if none was built for the given

include/clang/Driver/Options.td

 def fast : Flag<["-"], "fast">, Group<f_Group>;
 def fasynchronous_unwind_tables : Flag<["-"], "fasynchronous-unwind-tables">, Group<f_Group>;
 
 def fautolink : Flag <["-"], "fautolink">, Group<f_Group>;
 def fno_autolink : Flag <["-"], "fno-autolink">, Group<f_Group>,
   Flags<[DriverOption, CC1Option]>,
   HelpText<"Disable generation of linker directives for automatic library linking">;
 
+def fembed_bitcode : Flag<["-"], "fembed-bitcode">, Group<f_Group>,
+  Flags<[CC1Option, CC1AsOption]>,
+  HelpText<"Embed LLVM IR bitcode as data">;
+def fembed_bitcode_marker : Flag<["-"], "fembed-bitcode-marker">,
+  Group<f_Group>, Flags<[CC1Option]>,
+  HelpText<"Embed placeholder LLVM IR data as a marker">;
 def fgnu_inline_asm : Flag<["-"], "fgnu-inline-asm">, Group<f_Group>, Flags<[DriverOption]>;
 def fno_gnu_inline_asm : Flag<["-"], "fno-gnu-inline-asm">, Group<f_Group>,
   Flags<[DriverOption, CC1Option]>,
   HelpText<"Disable GNU style inline asm">;
 
 def fprofile_sample_use_EQ : Joined<["-"], "fprofile-sample-use=">,
     Group<f_Group>, Flags<[DriverOption, CC1Option]>,
     HelpText<"Enable sample-based profile guided optimizations">;

include/clang/Driver/ToolChain.h

virtual llvm::DebuggerKind getDefaultDebuggerTuning() const {
     return llvm::DebuggerKind::GDB;
   }
 
   /// UseSjLjExceptions - Does this tool chain use SjLj exceptions.
   virtual bool UseSjLjExceptions(const llvm::opt::ArgList &Args) const {
     return false;
   }
 
+  /// SupportsEmbeddedBitcode - Does this tool chain support embedded bitcode.
+  virtual bool SupportsEmbeddedBitcode() const {
+    return false;
+  }
+
   /// getThreadModel() - Which thread model does this target use?
   virtual std::string getThreadModel() const { return "posix"; }
 
   /// isThreadModelSupported() - Does this target support a thread model?
   virtual bool isThreadModelSupported(const StringRef Model) const;
 
   /// ComputeLLVMTriple - Return the LLVM target triple to use, after taking
   /// command line arguments into account.

lib/Driver/Driver.cpp

 using namespace clang::driver;
 using namespace clang;
 using namespace llvm::opt;
 
 Driver::Driver(StringRef ClangExecutable, StringRef DefaultTargetTriple,
                DiagnosticsEngine &Diags,
                IntrusiveRefCntPtr<vfs::FileSystem> VFS)
     : Opts(createDriverOptTable()), Diags(Diags), VFS(VFS), Mode(GCCMode),
-      SaveTemps(SaveTempsNone), LTOMode(LTOK_None),
+      SaveTemps(SaveTempsNone), BitcodeEmbed(EmbedNone), LTOMode(LTOK_None),
       ClangExecutable(ClangExecutable),
       SysRoot(DEFAULT_SYSROOT), UseStdLib(true),
       DefaultTargetTriple(DefaultTargetTriple),
       DriverTitle("clang LLVM compiler"), CCPrintOptionsFilename(nullptr),
       CCPrintHeadersFilename(nullptr), CCLogDiagnosticsFilename(nullptr),
       CCCPrintBindings(false), CCPrintHeaders(false), CCLogDiagnostics(false),
       CCGenDiagnostics(false), CCCGenericGCCName(""), CheckInputsExist(true),
       CCCUsePCH(true), SuppressMissingInputWarning(false) {
Compilation *Driver::BuildCompilation(ArrayRef<const char *> ArgList) {
 
   if (const Arg *A = Args.getLastArg(options::OPT_save_temps_EQ)) {
     SaveTemps = llvm::StringSwitch<SaveTempsMode>(A->getValue())
                     .Case("cwd", SaveTempsCwd)
                     .Case("obj", SaveTempsObj)
                     .Default(SaveTempsCwd);
   }
 
+  // Ignore -fembed-bitcode options with LTO
+  // since the output will be bitcode anyway.
+  if (!Args.hasFlag(options::OPT_flto, options::OPT_fno_lto, false)) {
+    if (Args.hasArg(options::OPT_fembed_bitcode))
+      BitcodeEmbed = EmbedBitcode;
+    else if (Args.hasArg(options::OPT_fembed_bitcode_marker))
+      BitcodeEmbed = EmbedMarker;
+  } else {
+    // claim the bitcode option under LTO so no warning is issued.
+    Args.ClaimAllArgs(options::OPT_fembed_bitcode);
+    Args.ClaimAllArgs(options::OPT_fembed_bitcode_marker);
+  }
+
   setLTOMode(Args);
 
   std::unique_ptr<llvm::opt::InputArgList> UArgs =
       llvm::make_unique<InputArgList>(std::move(Args));
 
   // Perform the default argument translations.
   DerivedArgList *TranslatedArgs = TranslateInputArgs(*UArgs);
 

 }
 
 // Returns a Tool for a given JobAction.  In case the action and its
 // predecessors can be combined, updates Inputs with the inputs of the
 // first combined action. If one of the collapsed actions is a
 // CudaHostAction, updates CollapsedCHA with the pointer to it so the
 // caller can deal with extra handling such action requires.
 static const Tool *selectToolForJob(Compilation &C, bool SaveTemps,
-                                    const ToolChain *TC, const JobAction *JA,
+                                    bool EmbedBitcode, const ToolChain *TC,
+                                    const JobAction *JA,
                                     const ActionList *&Inputs,
                                     const CudaHostAction *&CollapsedCHA) {
   const Tool *ToolForJob = nullptr;
   CollapsedCHA = nullptr;
 
   // See if we should look for a compiler with an integrated assembler. We match
   // bottom up, so what we are actually looking for is an assembler job with a
   // compiler input.
 
   if (TC->useIntegratedAs() && !SaveTemps &&
       !C.getArgs().hasArg(options::OPT_via_file_asm) &&
       !C.getArgs().hasArg(options::OPT__SLASH_FA) &&
       !C.getArgs().hasArg(options::OPT__SLASH_Fa) &&
       isa<AssembleJobAction>(JA) && Inputs->size() == 1 &&
       isa<BackendJobAction>(*Inputs->begin())) {
-    // A BackendJob is always preceded by a CompileJob, and without
-    // -save-temps they will always get combined together, so instead of
-    // checking the backend tool, check if the tool for the CompileJob
-    // has an integrated assembler.
+    // A BackendJob is always preceded by a CompileJob, and without -save-temps
+    // or -fembed-bitcode, they will always get combined together, so instead of
+    // checking the backend tool, check if the tool for the CompileJob has an
+    // integrated assembler. For -fembed-bitcode, CompileJob is still used to
+    // look up tools for BackendJob, but they need to match before we can split
+    // them.
     const ActionList *BackendInputs = &(*Inputs)[0]->getInputs();
     // Compile job may be wrapped in CudaHostAction, extract it if
     // that's the case and update CollapsedCHA if we combine phases.
     CudaHostAction *CHA = dyn_cast<CudaHostAction>(*BackendInputs->begin());
     JobAction *CompileJA = cast<CompileJobAction>(
         CHA ? *CHA->input_begin() : *BackendInputs->begin());
     assert(CompileJA && "Backend job is not preceeded by compile job.");
     const Tool *Compiler = TC->SelectTool(*CompileJA);
     if (!Compiler)
       return nullptr;
+    // When using -fembed-bitcode, it is required to have the same tool (clang)
+    // for both CompilerJA and BackendJA. Otherwise, combine two stages.
+    if (EmbedBitcode) {
+      JobAction *InputJA = cast<JobAction>(*Inputs->begin());
+      const Tool *BackendTool = TC->SelectTool(*InputJA);
+      if (BackendTool == Compiler)
+        CompileJA = InputJA;
+    }
     if (Compiler->hasIntegratedAssembler()) {
       Inputs = &CompileJA->getInputs();
       ToolForJob = Compiler;
       CollapsedCHA = CHA;
     }
   }
 
   // A backend job should always be combined with the preceding compile job
-  // unless OPT_save_temps is enabled and the compiler is capable of emitting
-  // LLVM IR as an intermediate output.
+  // unless OPT_save_temps or OPT_fembed_bitcode is enabled and the compiler is
+  // capable of emitting LLVM IR as an intermediate output.
   if (isa<BackendJobAction>(JA)) {
     // Check if the compiler supports emitting LLVM IR.
     assert(Inputs->size() == 1);
     // Compile job may be wrapped in CudaHostAction, extract it if
     // that's the case and update CollapsedCHA if we combine phases.
     CudaHostAction *CHA = dyn_cast<CudaHostAction>(*Inputs->begin());
     JobAction *CompileJA =
         cast<CompileJobAction>(CHA ? *CHA->input_begin() : *Inputs->begin());
     assert(CompileJA && "Backend job is not preceeded by compile job.");
     const Tool *Compiler = TC->SelectTool(*CompileJA);
     if (!Compiler)
       return nullptr;
-    if (!Compiler->canEmitIR() || !SaveTemps) {
+    if (!Compiler->canEmitIR() ||
+        (!SaveTemps && !EmbedBitcode)) {
       Inputs = &CompileJA->getInputs();
       ToolForJob = Compiler;
       CollapsedCHA = CHA;
     }
   }
 
   // Otherwise use the tool for the current job.
   if (!ToolForJob)
if (const CudaDeviceAction *CDA = dyn_cast<CudaDeviceAction>(A)) {
     return II;
   }
 
   const ActionList *Inputs = &A->getInputs();
 
   const JobAction *JA = cast<JobAction>(A);
   const CudaHostAction *CollapsedCHA = nullptr;
   const Tool *T =
-      selectToolForJob(C, isSaveTempsEnabled(), TC, JA, Inputs, CollapsedCHA);
+      selectToolForJob(C, isSaveTempsEnabled(), embedBitcodeEnabled(), TC, JA,
+                       Inputs, CollapsedCHA);
   if (!T)
     return InputInfo();
 
   // If we've collapsed action list that contained CudaHostAction we
   // need to build jobs for device-side inputs it may have held.
   if (CollapsedCHA) {
     for (const Action *DA : CollapsedCHA->getDeviceActions()) {
       CudaDeviceInputInfos.push_back(BuildJobsForAction(

lib/Driver/ToolChains.h

public:
   }
 
   bool SupportsObjCGC() const override;
 
   void CheckObjCARC() const override;
 
   bool UseSjLjExceptions(const llvm::opt::ArgList &Args) const override;
 
+  bool SupportsEmbeddedBitcode() const override;
+
   SanitizerMask getSupportedSanitizers() const override;
 };
 
 /// DarwinClang - The Darwin toolchain used by Clang.
 class LLVM_LIBRARY_VISIBILITY DarwinClang : public Darwin {
 public:
   DarwinClang(const Driver &D, const llvm::Triple &Triple,
               const llvm::opt::ArgList &Args);

lib/Driver/ToolChains.cpp

if (getTriple().getArch() != llvm::Triple::arm &&
       getTriple().getArch() != llvm::Triple::thumb)
     return false;
 
   // Only watchOS uses the new DWARF/Compact unwinding method.
   llvm::Triple Triple(ComputeLLVMTriple(Args));
   return !Triple.isWatchABI();
 }
 
+bool Darwin::SupportsEmbeddedBitcode() const {
+  assert(TargetInitialized && "Target not initialized!");
+  if (isTargetIPhoneOS() && isIPhoneOSVersionLT(6, 0))
+    return false;
+  return true;
+}
+
 bool MachO::isPICDefault() const { return true; }
 
 bool MachO::isPIEDefault() const { return false; }
 
 bool MachO::isPICDefaultForced() const {
   return (getArch() == llvm::Triple::x86_64 ||
           getArch() == llvm::Triple::aarch64);
 }

lib/Driver/Tools.cpp

void Clang::ConstructJob(Compilation &C, const JobAction &JA,
 
   if (const Arg *A = Args.getLastArg(options::OPT_fthinlto_index_EQ)) {
     if (!types::isLLVMIR(Input.getType()))
       D.Diag(diag::err_drv_argument_only_allowed_with) << A->getAsString(Args)
                                                        << "-x ir";
     Args.AddLastArg(CmdArgs, options::OPT_fthinlto_index_EQ);
   }
 
+  // Embed-bitcode option.
+  if (C.getDriver().embedBitcodeEnabled() &&
+      (isa<BackendJobAction>(JA) || isa<AssembleJobAction>(JA))) {
+    // Add flags implied by -fembed-bitcode.
+    CmdArgs.push_back("-fembed-bitcode");
+    // Disable all llvm IR level optimizations.
+    CmdArgs.push_back("-disable-llvm-optzns");
+  }
+  if (C.getDriver().embedBitcodeMarkerOnly())
+    CmdArgs.push_back("-fembed-bitcode-marker");
+
   // We normally speed up the clang process a bit by skipping destructors at
   // exit, but when we're generating diagnostics we can rely on some of the
   // cleanup.
   if (!C.isForDiagnostics())
     CmdArgs.push_back("-disable-free");
 
 // Disable the verification pass in -asserts builds.
 #ifdef NDEBUG
if (const Arg *A =
           Args.getLastArg(options::OPT_fpie, options::OPT_fPIE,
                           options::OPT_fno_pie, options::OPT_fno_PIE)) {
     if (A->getOption().matches(options::OPT_fpie) ||
         A->getOption().matches(options::OPT_fPIE))
       CmdArgs.push_back("-pie");
     else
       CmdArgs.push_back("-no_pie");
   }
+  // for embed-bitcode, use -bitcode_bundle in linker command
+  if (C.getDriver().embedBitcodeEnabled() ||
+      C.getDriver().embedBitcodeMarkerOnly()) {
+    // Check if the toolchain supports bitcode build flow.
+    if (MachOTC.SupportsEmbeddedBitcode())
+      CmdArgs.push_back("-bitcode_bundle");
+    else
+      D.Diag(diag::err_drv_bitcode_unsupported_on_toolchain);
+  }
 
   Args.AddLastArg(CmdArgs, options::OPT_prebind);
   Args.AddLastArg(CmdArgs, options::OPT_noprebind);
   Args.AddLastArg(CmdArgs, options::OPT_nofixprebinding);
   Args.AddLastArg(CmdArgs, options::OPT_prebind__all__twolevel__modules);
   Args.AddLastArg(CmdArgs, options::OPT_read__only__relocs);
   Args.AddAllArgs(CmdArgs, options::OPT_sectcreate);
   Args.AddAllArgs(CmdArgs, options::OPT_sectorder);

test/Driver/embed-bitcode.c

+// RUN: %clang -ccc-print-bindings -c %s -fembed-bitcode 2>&1 | FileCheck %s
+// CHECK: clang
+// CHECK: clang
+
+// RUN: %clang %s -fembed-bitcode 2>&1 -### | FileCheck %s -check-prefix=CHECK-CC
+// CHECK-CC: -cc1
+// CHECK-CC: -emit-llvm-bc
+// CHECK-CC: -cc1
+// CHECK-CC: -emit-obj
+// CHECK-CC: -fembed-bitcode
+// CHECK-CC: ld
+// CHECK-CC: -bitcode_bundle
+
+// RUN: %clang %s -save-temps -fembed-bitcode 2>&1 -### | FileCheck %s -check-prefix=CHECK-SAVE-TEMP
+// CHECK-SAVE-TEMP: -cc1
+// CHECK-SAVE-TEMP: -E
+// CHECK-SAVE-TEMP: -cc1
+// CHECK-SAVE-TEMP: -emit-llvm-bc
+// CHECK-SAVE-TEMP: -cc1
+// CHECK-SAVE-TEMP: -S
+// CHECK-SAVE-TEMP: -fembed-bitcode
+// CHECK-SAVE-TEMP: -cc1as
+// CHECK-SAVE-TEMP: ld
+// CHECK-SAVE-TEMP: -bitcode_bundle
+
+// RUN: %clang -c %s -flto -fembed-bitcode 2>&1 -### | FileCheck %s -check-prefix=CHECK-LTO
+// CHECK-LTO: -cc1
+// CHECK-LTO: -emit-llvm-bc
+// CHECK-LTO-NOT: warning: argument unused during compilation: '-fembed-bitcode'
+// CHECK-LTO-NOT: -cc1
+// CHECK-LTO-NOT: -fembed-bitcode
+
+// RUN: %clang -c %s -fembed-bitcode-marker 2>&1 -### | FileCheck %s -check-prefix=CHECK-MARKER
+// CHECK-MARKER: -cc1
+// CHECK-MARKER: -emit-obj
+// CHECK-MARKER: -fembed-bitcode-marker
+// CHECK-MARKER-NOT: -cc1
+