hama-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Edward J. Yoon" <edwardy...@apache.org>
Subject Re: 2D blocking algorithm
Date Tue, 21 Oct 2008 01:36:24 GMT
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

Mime
View raw message