Mailing-List: contact hama-dev-help@incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: hama-dev@incubator.apache.org
Received-SPF: pass (athena.apache.org: domain of guosijie@gmail.com designates
 209.85.128.187 as permitted sender)
DomainKey-Signature: a=rsa-sha1; c=nofws;
        d=gmail.com; s=gamma;
        h=message-id:date:from:to:subject:in-reply-to:mime-version
         :content-type:references;
        b=B2EdgfZjnPSYMOjA9G3q6JGNbTgdVWq66XLlGzNV7f6FNH1DvImbgtOYGmld4o8rPm
         EuDi4svCFeC7A4aIZGk4/+AV1xNS3b6H0jEd2f3QElosGREmOq/g4yu/aW6uhz3DyF3z
         VEKf9NyaSx8tIXGx2vI+uXy2rWldgurWNfR1c=
Message-ID: <25aacb800810201903n3bff1221kc0e17de2ca4dc259@mail.gmail.com>
Date: Tue, 21 Oct 2008 10:03:10 +0800
From: "Samuel Guo" <guosijie@gmail.com>
To: hama-dev@incubator.apache.org
Subject: Re: 2D blocking algorithm
In-Reply-To: <eb4706e0810201836i57d9b2d8r86827aeedf40b2e8@mail.gmail.com>
MIME-Version: 1.0
Content-Type: multipart/alternative;
	boundary="----=_Part_282_9657609.1224554590942"
References: <eb4706e0810201811h244aa9d3wf0a7978e5f7e908f@mail.gmail.com>
	 <25aacb800810201818k3573a0bdtf76da95dcc73871e@mail.gmail.com>
	 <eb4706e0810201836i57d9b2d8r86827aeedf40b2e8@mail.gmail.com>

------=_Part_282_9657609.1224554590942
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

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 <edwardyoon@apache.org>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 <guosijie@gmail.com> 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 <edwardyoon@apache.org>
> >
> >> 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
>

------=_Part_282_9657609.1224554590942--