rmaz added a comment. In D109632#3085187 <https://reviews.llvm.org/D109632#3085187>, @dexonsmith wrote:
> But another benefit of not double-storing transitively imported methods is > that it makes the PCMs more independent, tacking slightly closer to > "ImmediateDep1.pcm" being reproducible even when "SharedDep.pcm" adds a > method to the global pool. This is a nice property if it's not too expensive. > Looking at the numbers above, it doesn't look expensive; the relative > performance for @rmaz's motivating use case seems pretty small. > > @rmaz, will your goals be achieved by taking @vsapsai's approach? If so, I'm > leaning slightly that way. We can definitely work with it. From my perspective the faster solution would be preferred, but if there are potential future benefits from not storing dependent modules methods in a pcm them we can ship this solution and look at other possible avenues for performance improvement to gain parity with the non-modular case. 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
