On 12/01/2017 01:57 PM, Rebecca Cran wrote: > >> On Dec 1, 2017, at 1:50 PM, Jens Axboe <axboe@xxxxxxxxx> wrote: >> >> My goal would be to find if there's something simple we can do to provide >> a fairly optimized version we can use for larger copies, which is >> essentially just for mmap and libpmem/dev-dax and friends. > > Would you be looking for code that makes the decision at runtime > (using cpuid to see which of SSE, AVX, AVX512F it supports) or having > it be a compile-time thing? Should probably be a runtime thing, similar to how we do crc32c checks and enabling. And I suspect that we'll just end up with something that is just enabled for larger sizes, and/or just used for the performance sensitive engine copies. I don't think we have hot mem copies anywhere else. But we need to see a substantial win somewhere for it to be worth it. If not, then I'd rather not, since it does come with an extra maintenance burden. -- Jens Axboe -- To unsubscribe from this list: send the line "unsubscribe fio" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
