[Pw_forum] Single precision FFTs on GPUs for QE ?

[Konstantin Kudin]

 Hi all,

 With all the hype surrounding the recent graphics cards and how fast
they are supposed to be, the following question arises.

 If QE were to do everything in double precision, but the FFTs in
single precision, will it hurt too much? FFTs kind of do few
operations, so the error accumulation could be tolerable. Of course,
another issue if the few digits of single precision are really enough
...

 With multicore cpus it is increasingly difficult to transfer the data
between the boxes for FFTs, it sure would be easier (in principle) to
send the data locally to the super fast (tm) GPU.
 
 Anyway, from the implementation perspective, FFTs are currectly
reasonably well localized in the code, so in principle the hassle of
using GPUs would be only local in terms of coding :-) 

 Ideas? 

 Kostya


       
____________________________________________________________________________________Give spam the boot. Take control with tough spam protection in the all-new Yahoo! Mail Beta.
http://advision.webevents.yahoo.com/mailbeta/newmail_html.html