From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from zeniv.linux.org.uk ([195.92.253.2]:37494 "EHLO ZenIV.linux.org.uk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751765AbcGaBHG (ORCPT ); Sat, 30 Jul 2016 21:07:06 -0400 Date: Sun, 31 Jul 2016 00:52:15 +0100 From: Al Viro To: Linus Torvalds Cc: linux-fsdevel Subject: Re: [RFC] parent in ->d_compare() arguments Message-ID: <20160730235215.GA2356@ZenIV.linux.org.uk> References: <20160730010738.GY2356@ZenIV.linux.org.uk> <20160730233044.GZ2356@ZenIV.linux.org.uk> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: Sender: linux-fsdevel-owner@vger.kernel.org List-ID: On Sat, Jul 30, 2016 at 04:36:17PM -0700, Linus Torvalds wrote: > So? Aren't they in the same cacheline? Yes (unless lockdep map is stuck in between, in which case we are slow as hell anyway). > We've tried very hard to pack all those initial elements next to each other. > > The first-order approximation is that number of cacheline accesses > matter. And then the second order is to make code small and avoid > extra instructions. > > As far as I can tell, your change doesn't actually help the cacheline > accesses, and it makes the code bigger and have extra instructions. So > it doesn't appear to improve anything, and it does make things worse. > > But numbers talk, bullshit walks. If you have numbers to show > something different, that trumps my looking at code. I'll profile and post the results; not today, though - it's not urgent anyway, and right now I wouldn't trust my ability to do anything other than crawl to the bed and hopefully get some sleep (RDU -> BOS -> RDU, on top of 4 hours of sleep tonight and bloody two hours of delay in plane on the way back due to bad weather; picking the younger kid from summer STEM camp had been... interesting, for the lack of adequate printable words)