mboehme added a comment. In D128097#3605307 <https://reviews.llvm.org/D128097#3605307>, @nikic wrote:
> @mboehme Cycle counts are very noisy, and it's pretty much impossible to > determine whether they changed by looking at a single commit, unless the > differences are huge (like 10%). In this case, the commit got "lucky" and the > next commit goes back to the previous level. This is why I only look at > instruction counts. Thanks for the clarification! > And yes, there may be environment differences, either due to system headers, > or due to the used host compiler (which is GCC 9.3 for > llvm-compile-time-tracker). > > As this patch did mitigate half the regression, and the regression wasn't > particularly large in the first place, I personally wouldn't bother spending > too much time to investigate this further. In that case, I think I'll be content with the performance gains I've achieved here. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D128097/new/ https://reviews.llvm.org/D128097 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
