Reland "[ThinLTO][Bitcode] Generate import type in bitcode" (#97253)

https://github.com/llvm/llvm-project/pull/87600 was reverted in order to
revert
https://github.com/llvm/llvm-project/commit/6262763341fcd71a2b0708cf7485f9abd1d26ba8.
Now https://github.com/llvm/llvm-project/pull/95482 is fix forward for
https://github.com/llvm/llvm-project/commit/6262763341fcd71a2b0708cf7485f9abd1d26ba8.
This patch is a reland for
https://github.com/llvm/llvm-project/pull/87600

**Changes on top of original patch**
In `llvm/include/llvm/IR/ModuleSummaryIndex.h`, make the type of
`GVSummaryPtrSet` an `unordered_set` which is more memory efficient when
the number of elements is smaller than 128 [1]

**Original commit message**

For distributed ThinLTO, the LTO indexing step generates combined
summary for each module, and postlink pipeline reads the combined
summary which stores the information for link-time optimization.

This patch populates the 'import type' of a summary in bitcode, and
updates bitcode reader to parse the bit correctly.

[1]
https://github.com/llvm/llvm-project/blob/393eff4e02e7ab3d234d246a8d6912c8e745e6f9/llvm/lib/Support/SmallPtrSet.cpp#L43

1 files changed, 30 insertions, 8 deletions

diff --git a/llvm/lib/Bitcode/Writer/BitcodeWriter.cpp b/llvm/lib/Bitcode/Writer/BitcodeWriter.cpp
index 7a228fb..3378931 100644
--- a/llvm/lib/Bitcode/Writer/BitcodeWriter.cpp
+++ b/llvm/lib/Bitcode/Writer/BitcodeWriter.cpp
@@ -425,6 +425,11 @@ class IndexBitcodeWriter : public BitcodeWriterBase {
   /// The combined index to write to bitcode.
   const ModuleSummaryIndex &Index;
 
+  /// When writing combined summaries, provides the set of global value
+  /// summaries for which the value (function, function alias, etc) should be
+  /// imported as a declaration.
+  const GVSummaryPtrSet *DecSummaries = nullptr;
+
   /// When writing a subset of the index for distributed backends, client
   /// provides a map of modules to the corresponding GUIDs/summaries to write.
   const std::map<std::string, GVSummaryMapTy> *ModuleToSummariesForIndex;
@@ -453,11 +458,16 @@ public:
   /// Constructs a IndexBitcodeWriter object for the given combined index,
   /// writing to the provided \p Buffer. When writing a subset of the index
   /// for a distributed backend, provide a \p ModuleToSummariesForIndex map.
+  /// If provided, \p DecSummaries specifies the set of summaries for which
+  /// the corresponding functions or aliased functions should be imported as a
+  /// declaration (but not definition) for each module.
   IndexBitcodeWriter(BitstreamWriter &Stream, StringTableBuilder &StrtabBuilder,
                      const ModuleSummaryIndex &Index,
+                     const GVSummaryPtrSet *DecSummaries = nullptr,
                      const std::map<std::string, GVSummaryMapTy>
                          *ModuleToSummariesForIndex = nullptr)
       : BitcodeWriterBase(Stream, StrtabBuilder), Index(Index),
+        DecSummaries(DecSummaries),
         ModuleToSummariesForIndex(ModuleToSummariesForIndex) {
 
     // See if the StackIdIndex was already added to the StackId map and
@@ -1226,7 +1236,8 @@ static uint64_t getEncodedFFlags(FunctionSummary::FFlags Flags) {
 
 // Decode the flags for GlobalValue in the summary. See getDecodedGVSummaryFlags
 // in BitcodeReader.cpp.
-static uint64_t getEncodedGVSummaryFlags(GlobalValueSummary::GVFlags Flags) {
+static uint64_t getEncodedGVSummaryFlags(GlobalValueSummary::GVFlags Flags,
+                                         bool ImportAsDecl = false) {
   uint64_t RawFlags = 0;
 
   RawFlags |= Flags.NotEligibleToImport; // bool
@@ -1241,7 +1252,8 @@ static uint64_t getEncodedGVSummaryFlags(GlobalValueSummary::GVFlags Flags) {
 
   RawFlags |= (Flags.Visibility << 8); // 2 bits
 
-  RawFlags |= (Flags.ImportType << 10); // 1 bit
+  unsigned ImportType = Flags.ImportType | ImportAsDecl;
+  RawFlags |= (ImportType << 10); // 1 bit
 
   return RawFlags;
 }
@@ -4568,6 +4580,12 @@ void IndexBitcodeWriter::writeCombinedGlobalValueSummary() {
   Abbv->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::VBR, 8));
   unsigned AllocAbbrev = Stream.EmitAbbrev(std::move(Abbv));
 
+  auto shouldImportValueAsDecl = [&](GlobalValueSummary *GVS) -> bool {
+    if (DecSummaries == nullptr)
+      return false;
+    return DecSummaries->count(GVS);
+  };
+
   // The aliases are emitted as a post-pass, and will point to the value
   // id of the aliasee. Save them in a vector for post-processing.
   SmallVector<AliasSummary *, 64> Aliases;
@@ -4678,7 +4696,8 @@ void IndexBitcodeWriter::writeCombinedGlobalValueSummary() {
     NameVals.push_back(*ValueId);
     assert(ModuleIdMap.count(FS->modulePath()));
     NameVals.push_back(ModuleIdMap[FS->modulePath()]);
-    NameVals.push_back(getEncodedGVSummaryFlags(FS->flags()));
+    NameVals.push_back(
+        getEncodedGVSummaryFlags(FS->flags(), shouldImportValueAsDecl(FS)));
     NameVals.push_back(FS->instCount());
     NameVals.push_back(getEncodedFFlags(FS->fflags()));
     NameVals.push_back(FS->entryCount());
@@ -4727,7 +4746,8 @@ void IndexBitcodeWriter::writeCombinedGlobalValueSummary() {
     NameVals.push_back(AliasValueId);
     assert(ModuleIdMap.count(AS->modulePath()));
     NameVals.push_back(ModuleIdMap[AS->modulePath()]);
-    NameVals.push_back(getEncodedGVSummaryFlags(AS->flags()));
+    NameVals.push_back(
+        getEncodedGVSummaryFlags(AS->flags(), shouldImportValueAsDecl(AS)));
     auto AliaseeValueId = SummaryToValueIdMap[&AS->getAliasee()];
     assert(AliaseeValueId);
     NameVals.push_back(AliaseeValueId);
@@ -5068,8 +5088,9 @@ void BitcodeWriter::writeModule(const Module &M,
 
 void BitcodeWriter::writeIndex(
     const ModuleSummaryIndex *Index,
-    const std::map<std::string, GVSummaryMapTy> *ModuleToSummariesForIndex) {
-  IndexBitcodeWriter IndexWriter(*Stream, StrtabBuilder, *Index,
+    const std::map<std::string, GVSummaryMapTy> *ModuleToSummariesForIndex,
+    const GVSummaryPtrSet *DecSummaries) {
+  IndexBitcodeWriter IndexWriter(*Stream, StrtabBuilder, *Index, DecSummaries,
                                  ModuleToSummariesForIndex);
   IndexWriter.write();
 }
@@ -5124,12 +5145,13 @@ void IndexBitcodeWriter::write() {
 // index for a distributed backend, provide a \p ModuleToSummariesForIndex map.
 void llvm::writeIndexToFile(
     const ModuleSummaryIndex &Index, raw_ostream &Out,
-    const std::map<std::string, GVSummaryMapTy> *ModuleToSummariesForIndex) {
+    const std::map<std::string, GVSummaryMapTy> *ModuleToSummariesForIndex,
+    const GVSummaryPtrSet *DecSummaries) {
   SmallVector<char, 0> Buffer;
   Buffer.reserve(256 * 1024);
 
   BitcodeWriter Writer(Buffer);
-  Writer.writeIndex(&Index, ModuleToSummariesForIndex);
+  Writer.writeIndex(&Index, ModuleToSummariesForIndex, DecSummaries);
   Writer.writeStrtab();
 
   Out.write((char *)&Buffer.front(), Buffer.size());