[llvm] [memprof] Use uint32_t for linear call stack IDs (PR #93924)

[Teresa Johnson via llvm-commits]

================
@@ -259,6 +285,37 @@ static IndexedMemProfRecord deserializeV2(const MemProfSchema &Schema,
   return Record;
 }
 
+static IndexedMemProfRecord deserializeV3(const MemProfSchema &Schema,
+                                          const unsigned char *Ptr) {
+  using namespace support;
+
+  IndexedMemProfRecord Record;
+
+  // Read the meminfo nodes.
+  const uint64_t NumNodes =
+      endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
+  Record.AllocSites.reserve(NumNodes);
+  for (uint64_t I = 0; I < NumNodes; I++) {
+    IndexedAllocationInfo Node;
+    Node.CSId = endian::readNext<uint32_t, llvm::endianness::little>(Ptr);
+    Node.Info.deserialize(Schema, Ptr);
+    Ptr += PortableMemInfoBlock::serializedSize(Schema);
+    Record.AllocSites.push_back(Node);
+  }
+
+  // Read the callsite information.
+  const uint64_t NumCtxs =
+      endian::readNext<uint64_t, llvm::endianness::little>(Ptr);
+  Record.CallSiteIds.reserve(NumCtxs);
+  for (uint64_t J = 0; J < NumCtxs; J++) {
+    CallStackId CSId =
----------------
teresajohnson wrote:

CallStackId is still a typedef to uint64_t. I guess we can't change this completely while we still have v2 around, but it is a little confusing, and presumably using more memory than needed in the in-memory data structures. Should there be a TODO to switch this to 32-bits?

https://github.com/llvm/llvm-project/pull/93924