sammccall added a comment. My only concern with this is the overhead of the timing itself. We look up the hash bucket, turn timers on/off...
If I'm understanding right we do this once per AST node per registered matcher. Are you able to see any performance difference with this tracing on vs off? If there's a performance penalty but our goal is to gather enough statistics to be useful, we could consider only doing it every N reparses (or just with a certain probability). But even then, if it's slow maybe we should only be running it in cases where we know we're sending the stats somewhere useful. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D112835/new/ https://reviews.llvm.org/D112835 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
