Jennifer Schmitz <jschm...@nvidia.com> writes: >> It would also be good to check for performance regressions, now that we have >> a patch to test: >> I will run SPEC2017 with -mcpu=generic and -mcpu=native on Grace, but we >> would appreciate help with benchmarking on other platforms. >> Tamar, would you still be willing to test the patch on other platforms? >> >> If there are no other changes necessary and assuming there are no >> performance regressions, I was planning to commit the patch in January after >> returning from christmas break. >> >> In the meantime I wish everyone happy holidays. >> Jennifer > On Grace, the patch has no non-noise impact on performance for SPEC2017 with > -mcpu=generic and -mcpu=native. I also re-validated on aarch64 today, no > regression. > Do you advise to run additional performance tests or is the patch ready to be > pushed to trunk?
Go for it :) The patch is clearly moving in the right direction and it's a question of "when" not "if" we remove AARCH64_EXTRA_TUNE_USE_NEW_VECTOR_COSTS. We'll also get more performance coverage on non-aarch64 targets once the patch is pushed. Thanks again for doing this. Richard
