chandlerc wrote: > Something went wrong while rebasing (don't we have a policy to avoid force > pushing?), so the original changes are not included in this PR now...
Sorry, I wasn't aware of a policy, and the conflicts were stacking up so it seemed a bit simpler to rebase this onto top-of-tree and give a clean diff from there. If there is something I should do to make the changes more clear or help here, by all means. > [c-t-t stage2-clang > -0.99%](https://llvm-compile-time-tracker.com/compare.php?from=4bf956fe0e14aa7a6a91cad3feee99fd6eeea8f2&to=a459ae35edcbbbd1ad8d4caa20da389326d35af9&stat=instructions:u) > ([per-file > diff](https://llvm-compile-time-tracker.com/compare_clang.php?from=4bf956fe0e14aa7a6a91cad3feee99fd6eeea8f2&to=a459ae35edcbbbd1ad8d4caa20da389326d35af9&stat=instructions%3Au&sortBy=absolute-difference)). > Largely improvements, but there are some minor regressions as well, > primarily around clangDriver. Not sure what I should do with this... Very hard for me to tell if this is noise or something else, especially when the net savings are there. And to be clear, the savings for library-consumers of Clang are ... significant: 4-5% easily. So I would hope that is worth some small slowdown in specific driver files. Notably, the driver has at least never been the bottle neck and slowest compiles I have seen -- that is mostly in Sema. https://github.com/llvm/llvm-project/pull/216946 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
