[llvm] [CSSPGO] Compute and report profile matching recovered callsites and samples (PR #79090)

[via llvm-commits]

================
@@ -2443,53 +2372,222 @@ void SampleProfileMatcher::runOnFunction(const Function &F) {
   std::map<LineLocation, std::unordered_set<FunctionId>> ProfileAnchors;
   findProfileAnchors(*FSFlattened, ProfileAnchors);
 
-  // Detect profile mismatch for profile staleness metrics report.
-  // Skip reporting the metrics for imported functions.
-  if (!GlobalValue::isAvailableExternallyLinkage(F.getLinkage()) &&
-      (ReportProfileStaleness || PersistProfileStaleness)) {
-    // Use top-level nested FS for counting profile mismatch metrics since
-    // currently once a callsite is mismatched, all its children profiles are
-    // dropped.
-    if (const auto *FS = Reader.getSamplesFor(F))
-      countProfileMismatches(F, *FS, IRAnchors, ProfileAnchors);
-  }
+  // Compute the callsite match states for profile staleness report.
+  if (ReportProfileStaleness || PersistProfileStaleness)
+    computeCallsiteMatchStates(F, IRAnchors, ProfileAnchors, LocToLocMap());
 
   // Run profile matching for checksum mismatched profile, currently only
   // support for pseudo-probe.
   if (SalvageStaleProfile && FunctionSamples::ProfileIsProbeBased &&
       !ProbeManager->profileIsValid(F, *FSFlattened)) {
     // The matching result will be saved to IRToProfileLocationMap, create a new
     // map for each function.
+    auto &IRToProfileLocationMap = getIRToProfileLocationMap(F);
     runStaleProfileMatching(F, IRAnchors, ProfileAnchors,
-                            getIRToProfileLocationMap(F));
+                            IRToProfileLocationMap);
+    // Find and update callsite match states after matching.
+    if ((ReportProfileStaleness || PersistProfileStaleness) &&
+        !IRToProfileLocationMap.empty())
+      computeCallsiteMatchStates(F, IRAnchors, ProfileAnchors,
+                                 IRToProfileLocationMap);
   }
 }
 
-void SampleProfileMatcher::runOnModule() {
-  ProfileConverter::flattenProfile(Reader.getProfiles(), FlattenedProfiles,
-                                   FunctionSamples::ProfileIsCS);
-  for (auto &F : M) {
-    if (F.isDeclaration() || !F.hasFnAttribute("use-sample-profile"))
+void SampleProfileMatcher::computeCallsiteMatchStates(
+    const Function &F, const std::map<LineLocation, StringRef> &IRAnchors,
+    const std::map<LineLocation, std::unordered_set<FunctionId>>
+        &ProfileAnchors,
+    const LocToLocMap &IRToProfileLocationMap) {
+  // Use the matching result to determine if it's in post-match phrase.
+  bool IsPostMatch = !IRToProfileLocationMap.empty();
+  auto &MismatchedCallsites =
+      FuncCallsiteMatchStates[FunctionSamples::getCanonicalFnName(F.getName())];
+
+  auto MapIRLocToProfileLoc = [&](const LineLocation &IRLoc) {
+    const auto &ProfileLoc = IRToProfileLocationMap.find(IRLoc);
+    if (ProfileLoc != IRToProfileLocationMap.end())
+      return ProfileLoc->second;
+    else
+      return IRLoc;
+  };
+
+  std::set<LineLocation> MatchedCallsites;
+  for (const auto &I : IRAnchors) {
+    // In post-match, use the matching result to remap the current IR callsite.
+    const auto &Loc = MapIRLocToProfileLoc(I.first);
+    const auto &IRCalleeName = I.second;
+    const auto &It = ProfileAnchors.find(Loc);
+    if (It == ProfileAnchors.end())
       continue;
-    runOnFunction(F);
+    const auto &Callees = It->second;
+
+    // Since indirect call does not have CalleeName, check conservatively if
+    // callsite in the profile is a callsite location. This is to reduce num of
+    // false positive since otherwise all the indirect call samples will be
+    // reported as mismatching.
+    if (IRCalleeName == SampleProfileMatcher::UnknownIndirectCallee)
+      MatchedCallsites.insert(Loc);
+    // TODO : Ideally, we should ensure it's a direct callsite location(Callees
+    // size is 1). However, there may be a bug for profile merge(like ODR
+    // violation) that causes the callees size to be more than 1. After we fix
+    // the bug, we can remove this check.
+    else if (Callees.count(getRepInFormat(IRCalleeName)))
+      MatchedCallsites.insert(Loc);
+  }
+
+  // Check if there are any callsites in the profile that does not match to any
+  // IR callsites, those callsite samples will be discarded.
+  for (const auto &I : ProfileAnchors) {
+    const auto &Loc = I.first;
+    [[maybe_unused]] const auto &Callees = I.second;
+    assert(!Callees.empty() && "Callees should not be empty");
+    if (IsPostMatch) {
+      if (MatchedCallsites.count(Loc)) {
----------------
WenleiHe wrote:

Looking at how `MatchedCallsites` is being used, I'm thinking about whether we can hold everything directly in `MismatchedCallsites` and skip `MatchedCallsites`.

Specifically, how about we record the entries directly in `MismatchedCallsites` (with Recovered or Matched flag set depending on whether it's post-match) from the first loop. And then in this 2nd loop, we add in more new entries and mark these new entries Mismatched.

Additionally, I'm not sure `MismatchedCallsites` is a good name because it contains both matched and mismatched call sites. 

https://github.com/llvm/llvm-project/pull/79090