vsapsai added a comment. And for the empty module cache the numbers are
| **File** | **Baseline (s)** | **Set Dedupe (s)** | **No external (s)** | **Set Dedupe (percentage of baseline)** | **No external (percentage of baseline)** | | Project1. File1 | 14.080673 | 13.6296118 | 13.6954484 | 96.797% | 97.264% | | Project1. File2 | 15.2273466 | 14.6111798 | 14.7935872 | 95.954% | 97.151% | | Project1. File3 | 14.0047568 | 13.5988174 | 13.6155526 | 97.101% | 97.221% | | Project1. File4 | 13.0449874 | 12.5867316 | 12.8005844 | 96.487% | 98.126% | | Project1. File5 | 14.3756674 | 13.7928578 | 13.7562042 | 95.946% | 95.691% 💪 | | Though in this case the variance is much higher than before. Roughly it is 10 times higher and can reach 0.25. With such variance I think the difference between different fix approaches isn't meaningful and even comparing with the baseline is shaky. I haven't investigated further but suspect that the time of building modules dominates `addMethodToGlobalList` time and we cannot notice a meaningful difference. Methodology note: take average of 5 compilations instead of 8 as each compilation is longer. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D109632/new/ https://reviews.llvm.org/D109632 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits