Question about CULA's SVD before I spend money

General CULA Dense (LAPACK & BLAS) support and troubleshooting. Use this forum if you are having a general problem or have encountered a bug.

Question about CULA's SVD before I spend money

Postby whao » Tue Feb 07, 2012 3:07 pm

I just wanna understand the product better before I spend money/time on CUDA. Hope no one minds answering this question.

I'm curious as to the exact nature of the SVD function in CULA. Is it a specialized algorithm that uses all the processors to perform a single PCA, or is it just a version of SGESVD for each individual processor. I'm intending to do a lot of small SVDs. If it is the latter case, then I suspect I'll get working on a CUDA+CULA implementation right away :).
whao
 
Posts: 1
Joined: Mon Nov 07, 2011 11:27 am

Re: Question about CULA's SVD before I spend money

Postby kyle » Wed Feb 08, 2012 1:08 pm

Is it a specialized algorithm that uses all the processors to perform a single PCA, or is it just a version of SGESVD for each individual processor.

The implementation in CULA is used to process a single, typically large, PCA.

For more information about "burst" or "batch" modes, refer to some questions we've answered in others forum and blog posts.

Forum on burst Eigenvales: viewtopic.php?f=15&t=975
Blog on batch modes: http://www.culatools.com/blog/2011/12/0 ... perations/
kyle
Administrator
 
Posts: 301
Joined: Fri Jun 12, 2009 7:47 pm


Return to CULA Dense Support

Who is online

Users browsing this forum: Google [Bot] and 0 guests

cron