[clang][modules] Optimize construction and usage of the submodule index (#113391)

This patch avoids eagerly populating the submodule index on `Module`
construction. The `StringMap` allocation shows up in my profiles of
`clang-scan-deps`, while the index is not necessary most of the time. We
still construct it on-demand.

Moreover, this patch avoids performing qualified submodule lookup in
`ASTReader` whenever we're serializing a module graph whose top-level
module is unknown. This is pointless, since that's guaranteed to never
find any existing submodules anyway.

This speeds up `clang-scan-deps` by ~0.5% on my workload.

1 files changed, 7 insertions, 5 deletions

diff --git a/clang/lib/Basic/Module.cpp b/clang/lib/Basic/Module.cpp
index a7a3f6b..330108d 100644
--- a/clang/lib/Basic/Module.cpp
+++ b/clang/lib/Basic/Module.cpp
@@ -54,7 +54,6 @@ Module::Module(ModuleConstructorTag, StringRef Name,
     NoUndeclaredIncludes = Parent->NoUndeclaredIncludes;
     ModuleMapIsPrivate = Parent->ModuleMapIsPrivate;
 
-    Parent->SubModuleIndex[Name] = Parent->SubModules.size();
     Parent->SubModules.push_back(this);
   }
 }
@@ -351,11 +350,14 @@ void Module::markUnavailable(bool Unimportable) {
 }
 
 Module *Module::findSubmodule(StringRef Name) const {
-  llvm::StringMap<unsigned>::const_iterator Pos = SubModuleIndex.find(Name);
-  if (Pos == SubModuleIndex.end())
-    return nullptr;
+  // Add new submodules into the index.
+  for (unsigned I = SubModuleIndex.size(), E = SubModules.size(); I != E; ++I)
+    SubModuleIndex[SubModules[I]->Name] = I;
 
-  return SubModules[Pos->getValue()];
+  if (auto It = SubModuleIndex.find(Name); It != SubModuleIndex.end())
+    return SubModules[It->second];
+
+  return nullptr;
 }
 
 Module *Module::getGlobalModuleFragment() const {