sjoerdmeijer wrote: > Thanks for this PR. Do you have any compilation time and performance data?
This information is a bit spread out in the other tickets that I linked earlier, so to summarise that, compile times look really good and increases very minimal after the work that Madhur did. In https://github.com/llvm/llvm-project/pull/124911, I wrote: > The compile-time increase with a geomean increase of 0.19% looks good (after > committing https://github.com/llvm/llvm-project/pull/124247), I think: stage1-O3: Benchmark kimwitu++ +0.10% sqlite3 +0.14% consumer-typeset +0.07% Bullet +0.06% tramp3d-v4 +0.21% mafft +0.39% ClamAVi +0.06% lencod +0.61% SPASS +0.17% 7zip +0.08% geomean +0.19% Regarding performance, as I also wrote in that ticket, loop-interchange has a lot of potential. It triggers a lot of times e.g. in the LLVM test-suite, see this https://github.com/llvm/llvm-project/pull/124911#issuecomment-2624704156. It is now triggering slightly less than what I wrote in that comment because we made interchange more pessimistic to fix correctness issues, but we think that's okay because we consider getting interchange and DependenceAnalysis running by default as a first enablement step. Once we have achieved this, we are going to focus on performance and lift some of the restrictions (while maintaining correctness of course). With this first patch, interchange won't trigger on SPEC for example, but we plan to do that as follow up. https://github.com/llvm/llvm-project/pull/140182 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
