Return-Path: Delivered-To: apmail-incubator-hama-dev-archive@locus.apache.org Received: (qmail 62850 invoked from network); 21 Oct 2008 02:10:19 -0000 Received: from hermes.apache.org (HELO mail.apache.org) (140.211.11.2) by minotaur.apache.org with SMTP; 21 Oct 2008 02:10:19 -0000 Received: (qmail 85370 invoked by uid 500); 21 Oct 2008 02:10:21 -0000 Delivered-To: apmail-incubator-hama-dev-archive@incubator.apache.org Received: (qmail 85353 invoked by uid 500); 21 Oct 2008 02:10:21 -0000 Mailing-List: contact hama-dev-help@incubator.apache.org; run by ezmlm Precedence: bulk List-Help: List-Unsubscribe: List-Post: List-Id: Reply-To: hama-dev@incubator.apache.org Delivered-To: mailing list hama-dev@incubator.apache.org Received: (qmail 85332 invoked by uid 99); 21 Oct 2008 02:10:21 -0000 Received: from athena.apache.org (HELO athena.apache.org) (140.211.11.136) by apache.org (qpsmtpd/0.29) with ESMTP; Mon, 20 Oct 2008 19:10:21 -0700 X-ASF-Spam-Status: No, hits=1.2 required=10.0 tests=SPF_NEUTRAL X-Spam-Check-By: apache.org Received-SPF: neutral (athena.apache.org: local policy) Received: from [209.85.142.186] (HELO ti-out-0910.google.com) (209.85.142.186) by apache.org (qpsmtpd/0.29) with ESMTP; Tue, 21 Oct 2008 02:09:08 +0000 Received: by ti-out-0910.google.com with SMTP id w7so993361tib.6 for ; Mon, 20 Oct 2008 19:09:45 -0700 (PDT) Received: by 10.110.10.16 with SMTP id 16mr5482728tij.25.1224554985460; Mon, 20 Oct 2008 19:09:45 -0700 (PDT) Received: by 10.110.62.6 with HTTP; Mon, 20 Oct 2008 19:09:45 -0700 (PDT) Message-ID: Date: Tue, 21 Oct 2008 11:09:45 +0900 From: "Edward J. Yoon" Sender: edward@udanax.org To: hama-dev@incubator.apache.org Subject: Re: 2D blocking algorithm In-Reply-To: <25aacb800810201903n3bff1221kc0e17de2ca4dc259@mail.gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Content-Disposition: inline References: <25aacb800810201818k3573a0bdtf76da95dcc73871e@mail.gmail.com> <25aacb800810201903n3bff1221kc0e17de2ca4dc259@mail.gmail.com> X-Google-Sender-Auth: 99b7d30079d00f27 X-Virus-Checked: Checked by ClamAV on apache.org >> Is this block layout efficient in every matrix operation? >> How to choose the block size to make the parallel computation efficient? Good point. We should think/discuss about it. However, We have a capacity of 2d blocking at this time. :) On Tue, Oct 21, 2008 at 11:03 AM, Samuel Guo wrote: > hmm, your explanation is clear. > > but if we store the matrix to file, the block layout is fixed. > Is this block layout efficient in every matrix operation? > How to choose the block size to make the parallel computation efficient? > > On Tue, Oct 21, 2008 at 9:36 AM, Edward J. Yoon wrote: > >> For example, >> We have a 4 * 4 Matrix a : >> >> [a, b, c, d] >> [e, f, g, h] >> [i, j, k, l] >> [m, n, o, p]] >> >> Then, we can store an 2 * 2 block matrices to file. >> >> block(0, 0) is a 2 * 2 sub-matrix as below. >> >> [a, b] >> [e, f] >> >> Key : block(0, 0), Value : submatrix [a, b], [e, f] >> Key : block(0, 1), Value : submatrix [c, d], [g, h] >> Key : block(1, 0), Value : submatrix [i, j], [m, n] >> Key : block(1, 1), Value : submatrix [k, l], [o, p] >> >> ..... >> >> Also, If we can access sub-matrix from Hbase table, blocking can be >> pre-computed. >> >> On Tue, Oct 21, 2008 at 10:18 AM, Samuel Guo wrote: >> > should the matrices that do the multiplication have the same blockID >> layout? >> > >> > And How to divide a matrix into BlockIDs to make the parallel computation >> > most efficiently? >> > >> > 2008/10/21 Edward J. Yoon >> > >> >> Can anyone review this ?? >> >> >> >> http://blog.udanax.org/2008/10/parallel-matrix-multiply-on-hadoop.html >> >> >> >> -- >> >> Best regards, Edward J. Yoon >> >> edwardyoon@apache.org >> >> http://blog.udanax.org >> >> >> > >> >> >> >> -- >> Best regards, Edward J. Yoon >> edwardyoon@apache.org >> http://blog.udanax.org >> > -- Best regards, Edward J. Yoon edwardyoon@apache.org http://blog.udanax.org