[llvm-branch-commits] [llvm] [llvm-profgen] Support [buildid:]0xaddr format in perfscript input (PR #190863)

[Amir Ayupov via llvm-branch-commits]

https://github.com/aaupov updated https://github.com/llvm/llvm-project/pull/190863

>From a98cb95a1c3b46f242e6ffc30148d6c267ea9cf2 Mon Sep 17 00:00:00 2001
From: Amir Ayupov <aaupov at fb.com>
Date: Sun, 12 Apr 2026 12:15:02 -0700
Subject: [PATCH 1/2] reduce changes

Created using spr 1.3.4
---
 llvm/tools/llvm-profgen/PerfReader.cpp | 20 ++++++++++----------
 llvm/tools/llvm-profgen/PerfReader.h   |  3 +++
 2 files changed, 13 insertions(+), 10 deletions(-)

diff --git a/llvm/tools/llvm-profgen/PerfReader.cpp b/llvm/tools/llvm-profgen/PerfReader.cpp
index becabe4ab37bd..785946188b0b6 100644
--- a/llvm/tools/llvm-profgen/PerfReader.cpp
+++ b/llvm/tools/llvm-profgen/PerfReader.cpp
@@ -747,9 +747,9 @@ bool PerfScriptReader::extractLBRStack(TraceStream &TraceIt,
     // For pre-aggregated input, filter by build ID.
     if (IsPreAggregated) {
       StringRef BinaryBuildID = getFilterBuildID(Binary);
-      if (!SrcBuildID.empty() && SrcBuildID != BinaryBuildID)
+      if (SrcBuildID != BinaryBuildID)
         SrcIsInternal = false;
-      if (!DstBuildID.empty() && DstBuildID != BinaryBuildID)
+      if (DstBuildID != BinaryBuildID)
         DstIsInternal = false;
     }
     if (!SrcIsInternal)
@@ -775,8 +775,7 @@ bool PerfScriptReader::extractCallstack(TraceStream &TraceIt,
   // It's in bottom-up order with each frame in one line.
 
   // Extract stack frames from sample
-  while (!TraceIt.isAtEoF() &&
-         !isLBRSample(TraceIt.getCurrentLine(), true, IsPreAggregated)) {
+  while (!TraceIt.isAtEoF() && !isLBRSample(TraceIt.getCurrentLine(), true)) {
     StringRef FrameStr = TraceIt.getCurrentLine().ltrim();
     uint64_t FrameAddr = 0;
     StringRef FrameBuildID;
@@ -830,8 +829,7 @@ bool PerfScriptReader::extractCallstack(TraceStream &TraceIt,
   // Skip other unrelated line, find the next valid LBR line
   // Note that even for empty call stack, we should skip the address at the
   // bottom, otherwise the following pass may generate a truncated callstack
-  while (!TraceIt.isAtEoF() &&
-         !isLBRSample(TraceIt.getCurrentLine(), true, IsPreAggregated)) {
+  while (!TraceIt.isAtEoF() && !isLBRSample(TraceIt.getCurrentLine(), true)) {
     TraceIt.advance();
   }
   // Filter out broken stack sample. We may not have complete frame info
@@ -876,16 +874,14 @@ void HybridPerfReader::parseSample(TraceStream &TraceIt, uint64_t Count) {
   // Parsing call stack and populate into PerfSample.CallStack
   if (!extractCallstack(TraceIt, Sample->CallStack)) {
     // Skip the next LBR line matched current call stack
-    if (!TraceIt.isAtEoF() &&
-        isLBRSample(TraceIt.getCurrentLine(), true, IsPreAggregated))
+    if (!TraceIt.isAtEoF() && isLBRSample(TraceIt.getCurrentLine(), true))
       TraceIt.advance();
     return;
   }
 
   warnIfMissingMMap();
 
-  if (!TraceIt.isAtEoF() &&
-      isLBRSample(TraceIt.getCurrentLine(), true, IsPreAggregated)) {
+  if (!TraceIt.isAtEoF() && isLBRSample(TraceIt.getCurrentLine(), true)) {
     // Parsing LBR stack and populate into PerfSample.LBRStack
     if (extractLBRStack(TraceIt, Sample->LBRStack)) {
       if (IgnoreStackSamples) {
@@ -1227,6 +1223,10 @@ bool PerfScriptReader::isLBRSample(StringRef Line, bool CheckLineStart,
   return IsPreAggregated && Token.contains(":0x");
 }
 
+bool PerfScriptReader::isLBRSample(StringRef Line, bool CheckLineStart) {
+  return isLBRSample(Line, CheckLineStart, IsPreAggregated);
+}
+
 bool PerfScriptReader::isMMapEvent(StringRef Line) {
   // Short cut to avoid string find is possible.
   if (Line.empty() || Line.size() < 50)
diff --git a/llvm/tools/llvm-profgen/PerfReader.h b/llvm/tools/llvm-profgen/PerfReader.h
index f06c06d038a7b..7feb02b8c94c6 100644
--- a/llvm/tools/llvm-profgen/PerfReader.h
+++ b/llvm/tools/llvm-profgen/PerfReader.h
@@ -634,6 +634,9 @@ class PerfScriptReader : public PerfReaderBase {
 
   void setIsPreAggregated(bool V) { IsPreAggregated = V; }
 
+  // Check whether a given line is LBR sample
+  bool isLBRSample(StringRef Line, bool CheckLineStart);
+
 protected:
   // Check whether a given line is LBR sample
   static bool isLBRSample(StringRef Line, bool CheckLineStart,

>From b251df355cfab48e5be9185847a24a57e788f2bd Mon Sep 17 00:00:00 2001
From: Amir Ayupov <aaupov at fb.com>
Date: Fri, 1 May 2026 16:05:35 -0700
Subject: [PATCH 2/2] warn on mismatching buildid

Created using spr 1.3.4
---
 llvm/tools/llvm-profgen/PerfReader.cpp | 17 +++++++++++++----
 1 file changed, 13 insertions(+), 4 deletions(-)

diff --git a/llvm/tools/llvm-profgen/PerfReader.cpp b/llvm/tools/llvm-profgen/PerfReader.cpp
index b23e983116216..e08ba5ea96035 100644
--- a/llvm/tools/llvm-profgen/PerfReader.cpp
+++ b/llvm/tools/llvm-profgen/PerfReader.cpp
@@ -676,12 +676,21 @@ static bool parseAddress(StringRef Str, uint64_t &Addr, bool HasPrefix,
 }
 
 /// Return the build ID to use for filtering perfscript addresses.
-/// If --filter-build-id is specified, use it as an override.
+/// If --filter-build-id is specified, use it as an override (with a warning
+/// if it doesn't match the binary's auto-detected build ID).
 /// Otherwise, use the auto-detected value from the binary.
 static StringRef getFilterBuildID(const ProfiledBinary *Binary) {
-  if (FilterBuildID.getNumOccurrences() > 0)
-    return FilterBuildID;
-  return Binary->getFilterBuildID();
+  StringRef BinaryBuildID = Binary->getFilterBuildID();
+  if (FilterBuildID.getNumOccurrences() == 0)
+    return BinaryBuildID;
+  static bool Warned = false;
+  if (!Warned && !BinaryBuildID.empty() && FilterBuildID != BinaryBuildID) {
+    WithColor::warning() << "--filter-build-id=" << FilterBuildID
+                         << " does not match binary build ID " << BinaryBuildID
+                         << "\n";
+    Warned = true;
+  }
+  return FilterBuildID;
 }
 
 bool PerfScriptReader::extractLBRStack(TraceStream &TraceIt,