On 4/8/19 4:46 AM, Christoph Hellwig wrote: > Hi Jens, > > you complained about the overhead of calling nth_page in the multipage > bio_vec enabled I/O path a while ago. While I can't really reproduce > the numbers on my (slower) hardware we could avoid the nth_page calls > pretty easily with a few tweaks. Can you take a look at this series?
While I haven't had a chance to bench this yet, I like this series just as an improvement/cleanup. -- Jens Axboe
