Re: basic facts
you can pool 240*32 generators on the device, at any given
moment only 240 are executed. to compute the your performance
you would have to measure how long a single A5/1 computation takes
and multiply accordingly. Since you can get this number simply by
running the program, why bother?
240 cores should give you around 150million A5/1 per second at standard 600mhz.
> thank you .
> - so if my GPU says it has 240 Cores it means i can run 240*32 A5.1
> generator any given moment of time ? how much does it take for one A5.
> 1 to finish ? generally depending on my GPU configuration and its
> frequency how to compute my computation capability based on this code
> ?
>
> ---------- Forwarded message ----------
> From: *Sascha Krissler*<sascha.krissler@...>
> Date: Thu, Sep 3, 2009 at 11:44 PM
> Subject: Re: [A51] basic facts
> To: a51@...
>
> > here are my initial questions :
> >
> > - how many A5.1 engines can be running at the same time on a
> > Streaming Core with compute capability of 1.2 and 1.3 ?
>
> typical configuration is 1 block per MP and 256 threads per block
> resulting in 32threads per physical core.
>
> > - where to download the result of "occupancy test" excel file for
> > this implementation of A5.1 ?
>
> such a file does not exist.
>
> > - how much GPU memory we need for optimal operations on GTX285+ ?
>
> roughly 300kbyte. (32 bytes per chains being worked on times 30
> blocks times 256 threads)
> + 10-20 kbytes for the lookup tables.
>
> > - how to transfer data from GPU memory to main memory ? cudahostalloc
>
> > ?
>
> cudaMemcpyDeviceToHost
>
> ________________________________________________________________
>
> _______________________________________________ A51 mailing list A51 <at>
> lists.reflextor.com http://lists.lists.reflextor.com/cgi-bin/mailman/
>
> listinfo/a51
________________________________________________________________
Neu: WEB.DE Doppel-FLAT mit Internet-Flatrate + Telefon-Flatrate
für nur 19,99 Euro/mtl.!* http://produkte.web.de/go/02/