[llvm] [Bolt] Add a new hidden option to perf2bolt for testing purpose (PR #163785)

[Ádám Kallai via llvm-commits]

================
@@ -372,6 +379,80 @@ void DataAggregator::parsePreAggregated() {
   }
 }
 
+bool DataAggregator::isMMapEvent(StringRef Line) {
+  // Short cut to avoid string find is possible.
+  if (Line.empty() || Line.size() < 50)
+    return false;
+
+  // Check that PERF_RECORD_MMAP2 or PERF_RECORD_MMAP appear in the line.
+  return Line.contains("PERF_RECORD_MMAP");
+}
+
+void DataAggregator::parsePerfScriptEvents() {
+  outs() << "PERF2BOLT: parsing a hybrid perf-script events...\n";
+  NamedRegionTimer T("parsePerfScriptEvents", "Parsing perf-script events",
+                     TimerGroupName, TimerGroupDesc, opts::TimeAggregator);
+
+  ErrorOr<std::unique_ptr<MemoryBuffer>> MB =
+      MemoryBuffer::getFileOrSTDIN(Filename);
+  if (std::error_code EC = MB.getError()) {
+    errs() << "PERF2BOLT-ERROR: cannot open " << Filename << ": "
+           << EC.message() << "\n";
+    exit(1);
+  }
+
+  FileBuf = std::move(*MB);
+  ParsingBuf = FileBuf->getBuffer();
+  Col = 0;
+  Line = 1;
+  std::string MMapEvents = "";
+  std::string BranchEvents = "";
----------------
kaadam wrote:

Thanks for your feedback.
Yes, that was my first idea. However using the following `perf script --show-mmap-events -F pid,brstack ...` command by default to extract 'mmap' and 'brstack' events for `brstack` aggregation, I found sometimes the output is not exactly an ordered list. I mean an `mmap` event might appear among the branch events. That's why I moved to create a pre-selection logic before we call the proper functions. I agree it is costly. 

All in all if we can guarantee the strict order of the entries (as a prerequisites), that case we can call `parseMMapEvents` first, and when the first parse error occurs we can invoke `parseBranchEvents` on the Buffer as you suggested.

Side note: Using `parseMMapEvents` directly on the parsing buffer works correctly, the mmap function is able to handle other events like `brstack`. However if `parseBranchEvents` finds an mmap event, it will report an error. 

Currently I focused on branch event aggregation only, but I agree I need to cover other events and aggregation types. 

https://github.com/llvm/llvm-project/pull/163785