[NFC][ThinLTO] Let llvm::EmbedBitcodeInModule handle serialization.

[mtrofin]

Summary

llvm::EmbedBitcodeInModule handles serializing the passed-in module, if
the provided MemoryBufferRef is invalid. This is already the path taken
in one of the uses of the API - clang::EmbedBitcode, when called from
BackendConsumer::HandleTranslationUnit - so might as well do the same
here and reduce (by very little) code duplication.

The only difference this patch introduces is that the serialization happens
with ShouldPreserveUseListOrder set to true.

[tejohnson]

Does this rely on the following check in EmbedBitcodeInModule returning true?

if (!isBitcode((const unsigned char *)Buf.getBufferStart(),
               (const unsigned char *)Buf.getBufferEnd())) {

I.e. I assume an empty buffer will be !isBitcode? The comments there suggest this is trying to detect whether we have bitcode vs LLVM assembly as input. Might be better to add an explicit check for an empty buffer and expand the comments (along with your current change).

[mtrofin]

comments

[mtrofin]

explicit check

[tejohnson]

LGTM with comment updates as suggested below.

llvm/include/llvm/Bitcode/BitcodeWriter.h

159

maybe make this something like "...if the provided Buf is not bitcode (e.g. is LLVM assembly)" since that seems to be what it is trying to handle

llvm/lib/Bitcode/Writer/BitcodeWriter.cpp

4837

Update comment.

[mtrofin]

comments

llvm/lib/Bitcode/Writer/BitcodeWriter.cpp

4837

none of that comment applies, really

llvm/include/llvm/Bitcode/BitcodeWriter.h

class raw_ostream;
   /// writing the combined index file for ThinLTO. When writing a subset of the
   /// index for a distributed backend, provide the \p ModuleToSummariesForIndex
   /// map.
   void WriteIndexToFile(const ModuleSummaryIndex &Index, raw_ostream &Out,
                         const std::map<std::string, GVSummaryMapTy>
                             *ModuleToSummariesForIndex = nullptr);
 
   /// Save a copy of the llvm IR as data in the __LLVM,__bitcode section.
+  /// If available, pass the serialized module via the Buf parameter. If not,
+  /// pass an empty (default-initialized) MemoryBufferRef, and the serialization
+  /// will be handled by this API. The same behavior happens if the provided Buf
+  /// is not bitcode (i.e. if it's invalid data or even textual LLVM assembly).
   void EmbedBitcodeInModule(Module &M, MemoryBufferRef Buf, bool EmbedBitcode,
                             bool EmbedMarker,
                             const std::vector<uint8_t> *CmdArgs);
 
 } // end namespace llvm
 
 #endif // LLVM_BITCODE_BITCODEWRITER_H

llvm/lib/Bitcode/Writer/BitcodeWriter.cpp

void llvm::EmbedBitcodeInModule(llvm::Module &M, llvm::MemoryBufferRef Buf,
   }
   if (Used)
     Used->eraseFromParent();
 
   // Embed the bitcode for the llvm module.
   std::string Data;
   ArrayRef<uint8_t> ModuleData;
   Triple T(M.getTargetTriple());
-  // Create a constant that contains the bitcode.
-  // In case of embedding a marker, ignore the input Buf and use the empty
-  // ArrayRef. It is also legal to create a bitcode marker even Buf is empty.
+
   if (EmbedBitcode) {
-    if (!isBitcode((const unsigned char *)Buf.getBufferStart(),
+    if (Buf.getBufferSize() == 0 ||
+        !isBitcode((const unsigned char *)Buf.getBufferStart(),
                    (const unsigned char *)Buf.getBufferEnd())) {
       // If the input is LLVM Assembly, bitcode is produced by serializing
       // the module. Use-lists order need to be preserved in this case.
       llvm::raw_string_ostream OS(Data);
       llvm::WriteBitcodeToFile(M, OS, /* ShouldPreserveUseListOrder */ true);
       ModuleData =
           ArrayRef<uint8_t>((const uint8_t *)OS.str().data(), OS.str().size());
     } else
       // If the input is LLVM bitcode, write the input byte stream directly.
       ModuleData = ArrayRef<uint8_t>((const uint8_t *)Buf.getBufferStart(),

llvm/lib/LTO/LTOBackend.cpp

 
 static cl::opt<bool> EmbedBitcode(
     "lto-embed-bitcode", cl::init(false),
     cl::desc("Embed LLVM bitcode in object files produced by LTO"));
 
 static void EmitBitcodeSection(Module &M) {
   if (!EmbedBitcode)
     return;
-  SmallVector<char, 0> Buffer;
-  raw_svector_ostream OS(Buffer);
-  WriteBitcodeToFile(M, OS);
-
-  std::unique_ptr<MemoryBuffer> Buf(
-      new SmallVectorMemoryBuffer(std::move(Buffer)));
-  llvm::EmbedBitcodeInModule(M, Buf->getMemBufferRef(), /*EmbedBitcode*/ true,
+  llvm::EmbedBitcodeInModule(M, llvm::MemoryBufferRef(), /*EmbedBitcode*/ true,
                              /*EmbedMarker*/ false, /*CmdArgs*/ nullptr);
 }
 
 void codegen(const Config &Conf, TargetMachine *TM, AddStreamFn AddStream,
              unsigned Task, Module &Mod,
              const ModuleSummaryIndex &CombinedIndex) {
   if (Conf.PreCodeGenModuleHook && !Conf.PreCodeGenModuleHook(Task, Mod))
     return;