vsk added subscribers: mzolotukhin, vsk. vsk added a comment. Thanks for working on this. Collecting better timing information in the frontend sgtm. It's cheap to do, and we can use the information to guide our efforts re: attacking the compile-time problem. Feel free to add me to future timing-related reviews. Regarding this specific patch:
Could you add a short test (perhaps in test/Misc?) that checks that -ftime-report prints out an entry for 'Preprocessing'? It'd be nice to dump this timer from Preprocessor::PrintStats(), too. ================ Comment at: lib/Lex/Preprocessor.cpp:660 + llvm::TimeRegion(PPOpts->ShowTimers ? &PreprocessingTimer : nullptr); + ---------------- I wonder whether this is too fine-grained. I think setting up a timer in Preprocessor::Lex() might capture more information. Would you mind experimenting with that? ================ Comment at: lib/Lex/Preprocessor.cpp:660 + llvm::TimeRegion(PPOpts->ShowTimers ? &PreprocessingTimer : nullptr); + ---------------- vsk wrote: > I wonder whether this is too fine-grained. I think setting up a timer in > Preprocessor::Lex() might capture more information. Would you mind > experimenting with that? Nitpick: it may be useful to add PPOpts::getTimer(), in case we find more sites where we need to either get back the PP timer or nullptr. https://reviews.llvm.org/D36492 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
