From mboxrd@z Thu Jan 1 00:00:00 1970 From: "H. Peter Anvin" Subject: Re: [PATCH 06/13] async_tx: add support for asynchronous GF multiplication Date: Fri, 20 Mar 2009 16:25:41 -0700 Message-ID: <49C425F5.8060309@intel.com> References: <20090318191248.20375.40560.stgit@dwillia2-linux.ch.intel.com> <20090318192046.20375.89854.stgit@dwillia2-linux.ch.intel.com> <49C26D91.10804@intel.com> <49C41C1E.4060401@zytor.com> <49C4201A.5060905@emcraft.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <49C4201A.5060905@emcraft.com> Sender: linux-raid-owner@vger.kernel.org To: Ilya Yanok Cc: "H. Peter Anvin" , "Williams, Dan J" , "linux-raid@vger.kernel.org" , "linux-kernel@vger.kernel.org" , "neilb@suse.de" , "Sosnowski, Maciej" , Yuri Tikhonov List-Id: linux-raid.ids Ilya Yanok wrote: >>> >> That doesn't explain what you expect the code to do. It still doesn't >> make sense to me. > > We defined async_pq() function as hardware can do GF multiplication of > arbitrary elements and we want to take advantage of this fact. But by > the design of ASYNC_TX API we have to provide synchronous version of > that function too. > I don't think that makes sense since the algorithm doesn't use vector-vector multiplications and there is no need for them. As such, you're actively excluding support for hardware which only provides scalar-vector multiplication, even though that is all we need. In fact, vectorizing the constant doesn't make much sense. -hpa