I think it is better to let blocksize to be configured. even, the blocksize may be fit with the cpu L1 cache (1M, or 2M) which will make the in-memory block computation(addtion, multiplication) .
On Wed, Nov 12, 2008 at 7:57 PM, Edward J. Yoon <[EMAIL PROTECTED]>wrote: > Oh, Problem is a block size. BlockedDenseMatrix seems better. > > On Wed, Nov 12, 2008 at 8:46 PM, Edward J. Yoon <[EMAIL PROTECTED]> > wrote: > > To handle large matrix, maybe we should block-level store the matrix > > as a blocked dense matrix. Of course, the more memory will be used for > > get/set (i, j). > > > > -- > > Best Regards, Edward J. Yoon @ NHN, corp. > > [EMAIL PROTECTED] > > http://blog.udanax.org > > > > > > -- > Best Regards, Edward J. Yoon @ NHN, corp. > [EMAIL PROTECTED] > http://blog.udanax.org >
