From: Arnd Bergmann Subject: Re: [RFC] Alignment of fields in struct dentry Date: Mon, 18 Sep 2006 23:54:04 +0200 Message-ID: <200609182354.04781.arnd@arndb.de> References: <20060914093123.GA10431@wohnheim.fh-wedel.de> <200609152244.07889.arnd@arndb.de> <20060918212423.GB6899@wohnheim.fh-wedel.de> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org Return-path: Received: from moutng.kundenserver.de ([212.227.126.171]:52207 "EHLO moutng.kundenserver.de") by vger.kernel.org with ESMTP id S1751117AbWIRVyK convert rfc822-to-8bit (ORCPT ); Mon, 18 Sep 2006 17:54:10 -0400 To: =?iso-8859-1?q?J=F6rn_Engel?= In-Reply-To: <20060918212423.GB6899@wohnheim.fh-wedel.de> Content-Disposition: inline Sender: linux-fsdevel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On Monday 18 September 2006 23:24, J=F6rn Engel wrote: > On Fri, 15 September 2006 22:44:07 +0200, Arnd Bergmann wrote: > > > > I'd guess that a 32 byte alignment is much better here, 64 byte sou= nds > > excessive. It should have the same effect with the current dentry l= ayout > > and default config options, but would keep the d_iname length in th= e > > 16-44 byte range instead of 16-76 byte as your patch does. > >=20 > > Since all important fields are supposed to be kept in 32 bytes anyw= ay, > > they are still either at the start or the end of a given cache line= , > > but never cross two. >=20 > Another take would be to use a cacheline. But I guess the difference > between 32/64/cacheline is mostly academic, given the rate of changes > to struct dentry. There have been so many optimizations and misoptimizations regarding the dentry struct over the years. See http://lkml.org/lkml/2004/5/8/117 for the almost exact opposite of this patch, along with the same discus= sion that we're having now. Arnd <>< - To unsubscribe from this list: send the line "unsubscribe linux-fsdevel= " in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html