On 12/6/18 12:38 PM, Jeff Moyer wrote: > Jens Axboe <axboe@xxxxxxxxx> writes: > >> On 12/6/18 12:27 PM, Jeff Moyer wrote: >>> Jens Axboe <axboe@xxxxxxxxx> writes: >>> >>>> It's 192 bytes, fairly substantial. Most items don't need to be cleared, >>>> especially not upfront. Clear the ones we do need to clear, and leave >>>> the other ones for setup when the iocb is prepared and submitted. >>> >>> What performance gains do you see from this? >> >> Before this, I had 1% in memset doing high IOPS. With it, that's gone. >> 1% is a lot, when you have just one thread doing everything from submission >> to completion. > > I'm used to customers complaining about fractions of a percent, so I get > it. :-) I just wanted to know we had some measurable impact, as I've > seen bugs crop up from code like this in the past. Oh for sure, I wouldn't do it if there wasn't a noticeable gain from this! -- Jens Axboe
