From mboxrd@z Thu Jan 1 00:00:00 1970 From: "H. Peter Anvin" Subject: Re: with raid-6 any writes access all disks Date: Tue, 01 Nov 2011 15:22:44 -0700 Message-ID: <4EB07134.9080308@zytor.com> References: <20111027082331.01e1fc7a@notabene.brown> <4EA889FF.90002@zytor.com> <4EA92493.1060107@westcontrol.com> <4EA94D1F.8080507@zytor.com> <4EA956FE.3070504@westcontrol.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <4EA956FE.3070504@westcontrol.com> Sender: linux-raid-owner@vger.kernel.org To: David Brown Cc: NeilBrown , Chris Pearson , linux-raid@vger.kernel.org List-Id: linux-raid.ids On 10/27/2011 06:05 AM, David Brown wrote: > > Where does it go wrong? Is it the automatic vectorisation with SSE, > etc., that is still too limited with gcc? I have done very little work > with x86/amd64 assembly (most of my experience is with microcontrollers > rather than "big" processors), so I haven't tried looking at gcc's SSE > code and comparing it to hand-optimised code. > The autovectorization isn't good enough to understand the tricks that are necessary to get good performance. They require leaning pretty hard on the instruction set. -hpa