[MemProf] Switch to DenseMap for performance (NFC)

[teresajohnson]

Some profiling showed that the accesses to this map during bitcode
reading was incurring over 10% of the time in a large thin link. There
is no need for it to be std::map, and I measured around 8.5% time
reduction in the same thin link from switching to DenseMap.

[llvmbot]

@llvm/pr-subscribers-llvm-ir

Author: Teresa Johnson (teresajohnson)

Changes

Some profiling showed that the accesses to this map during bitcode
reading was incurring over 10% of the time in a large thin link. There
is no need for it to be std::map, and I measured around 8.5% time
reduction in the same thin link from switching to DenseMap.

---

Full diff: https://github.com/llvm/llvm-project/pull/81035.diff

1 Files Affected:

• (modified) llvm/include/llvm/IR/ModuleSummaryIndex.h (+1-1)

diff --git a/llvm/include/llvm/IR/ModuleSummaryIndex.h b/llvm/include/llvm/IR/ModuleSummaryIndex.h
index 395c9e332eb8f..d5ff15063671d 100644
--- a/llvm/include/llvm/IR/ModuleSummaryIndex.h
+++ b/llvm/include/llvm/IR/ModuleSummaryIndex.h
@@ -1377,7 +1377,7 @@ class ModuleSummaryIndex {
 
   // Temporary map while building StackIds list. Clear when index is completely
   // built via releaseTemporaryMemory.
-  std::map<uint64_t, unsigned> StackIdToIndex;
+  DenseMap<uint64_t, unsigned> StackIdToIndex;
 
   // YAML I/O support.
   friend yaml::MappingTraits<ModuleSummaryIndex>;