Comments for Where CFD meets GPU http://vratis.com/blog Just another WordPress site Wed, 15 Feb 2012 13:48:18 +0000 hourly 1 http://wordpress.org/?v=3.2.1 Comment on Performance of SpMV in CUSPARSE, CUSP and SpeedIT by Jakub Pola http://vratis.com/blog/?p=1#comment-43 Jakub Pola Wed, 15 Feb 2012 13:48:18 +0000 http://vratis.com/blog/?p=1#comment-43 Dear Klaus, Thank you very much for your suggestions but unfortunately that is not included in our road map. Dear Klaus,

Thank you very much for your suggestions but unfortunately that is not included in our road map.

]]> Comment on Performance of SpMV in CUSPARSE, CUSP and SpeedIT by Klaus http://vratis.com/blog/?p=1#comment-34 Klaus Sun, 05 Feb 2012 17:52:06 +0000 http://vratis.com/blog/?p=1#comment-34 Do you consider implementing a mixed precision version to leverage the single precision performance of current GPUs for double precision problems. An approach is described in the following paper: http://digbib.ubka.uni-karlsruhe.de/volltexte/1000019902 Some papers show, that by applying this approach a GTX 580 outperforms a Tesla C2050 GPU when solving double precision problems. Do you consider implementing a mixed precision version to leverage the single precision performance of current GPUs for double precision problems.

An approach is described in the following paper:
http://digbib.ubka.uni-karlsruhe.de/volltexte/1000019902

Some papers show, that by applying this approach a GTX 580 outperforms a Tesla C2050 GPU when solving double precision problems.

]]>