Hi, > Thanks for that great e-mail.
thanks for your interest ;-) > I'd be glad if you could send me the randomized SVD paper. I'll take a > look at this weekend. With the 'standard' SVD yes, I'll take a look at > the OpenCL part. For the generalized SVD is there an implementation for > N=2 in ViennaCL? I just recalled that the slides I was referring to are actually available online: http://web.stanford.edu/group/mmds/slides2010/Martinsson.pdf As you can see from page 9, the core part consists of just a few steps. And it's only a rank-k approximate SVD instead of a full SVD, but this is often sufficient (and faster...). The random numbers need to be generated on the host, we don't have random number generation on the GPU. Fortunately this is not expected to be a performance bottleneck, though... > I did already fork both viennacl-dev and PyViennaCl-dev. I will re-fork > it on Monday and start from there. Also, I'm comfortable with git, so > pull-requests through that should not be a problem. Excellent. :-) Best regards, Karli ------------------------------------------------------------------------------ _______________________________________________ ViennaCL-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/viennacl-devel
