[gpaw-users] rpa gpu on Emerald

[Marcin Dulak]

Hi,

On 11/19/2014 11:30 AM, Glenn Jones wrote:
>
> Hi Marcin,
>
> I have battled with various libraries and compilers and have managed 
> to get the rpa-gpu branch you sent me to compile using some of the 
> standard modules available on Emerald.
>
> Whether this is an efficient, or successful compilation when running 
> jobs I now need to check.
>
> Do you have some suitable scripts/protocols I could use as test jobs 
> to check the efficiency. I am going to try the example off the wiki 
> for benzene on metal surfaces – but this may be too long for initial 
> tests.
>
this is the test i know:
https://trac.fysik.dtu.dk/projects/gpaw/browser/branches/rpa-gpu-expt/gpaw/test/rpa_Na1_cuda.py
Note that only 1 GPU can be used together with 1 CPU core,
but there is no binding to GPU implemented in the code.
If you run in a queue and share GPU node with other users then this may 
be a problem as the code will always
try to use nvidia device 0.

Best regards,

Marcin
>
> Best regards,
>
> Glenn
>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.fysik.dtu.dk/pipermail/gpaw-users/attachments/20141119/bdec439c/attachment.html>