incubator-hama-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Samuel Guo" <guosi...@gmail.com>
Subject Re: 2D blocking algorithm
Date Thu, 23 Oct 2008 08:53:45 GMT
great idea.

On Thu, Oct 23, 2008 at 4:29 PM, Edward J. Yoon <edwardyoon@apache.org>wrote:

> Oh... one more...
>
> In Hbase, each cell can be sub-matrix.
>
> On Tue, Oct 21, 2008 at 11:09 AM, Edward J. Yoon <edwardyoon@apache.org>
> wrote:
> >>> 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 <guosijie@gmail.com> 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 <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
> >>>
> >>
> >
> >
> >
> > --
> > 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
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message