On Fri, Nov 13, 2009 at 01:57:04PM +0100, Frank A. Stevenson wrote: > In reply to questions on what hardware to buy for this particular > project, ATI seems to have a speed advantage at the moment. I took > Saschas ideas (thanks) on bitslicing A5/1 - and made an ATI implementation.
Thats some great news. I will try my best to make nvidia take the lead again. And the ATI code will find its way into the framework real soon now. (but you can run Franks version on its own too) > > A standards ATI Radon 5850 card will now deliver > 400 chains / second > at a sustained rate. Meaning a single table can be completed in 11 days. > BTW why do we say we need 2^8.5 tables, and not 360 which is much more > understandable to the layperson. It would be nice to have a simple > progress statistics showing how many tables have been completed of the > 360 instead of mysterious graphs with tiny writing along the axis. the code that processes the stats is in svn:/server/. status.cgi simply records all parameters of the client status report and writes them to a file and there is not much needed to improve it. process-stats.pl is the surviving quick hack that generates the graphs and it is probably a good idea to start it from scratch. Much of the quirkiness comes from the fact that clients did not report the number of chains sent since last report and the code tried to calculate it from the total chain count. The October 25 version does send that value. > > Code will be made available (probably through svn) > Thanks for the contribution, Frank. _______________________________________________ A51 mailing list [email protected] http://lists.lists.reflextor.com/cgi-bin/mailman/listinfo/a51
