Dear Fillipo, 

Great! I followed your instructions (--disable-parallel and --disable-magma) 
and it is working perfectly. Regarding the system under study, it is the 
example I took from "/espresso/NEB/example" directory and was, in a first 
moment, used only for test. The "real" system benchmark is running right now. 

By the way, as a beginner in GPU world, I still playing with the tools... I'm 
using a "zotac gtx 460". If I'm not wrong this is a GPU with 2.0 capability, so 
I though 20 was the correct arch. 

My apologies for subscribe to the wrong list. The next messages will be 
adressed to http://qe-forge.org/mailman/listinfo/q-e-gpgpu . 

Thank you very much for your time. 

Best regards, 
Eugenio 

----- Mensagem original ----- 
De: "Filippo Spiga" <spiga.filippo at gmail.com> 
Para: "PWSCF Forum" <pw_forum at pwscf.org> 
Cc: q-e-gpgpu at qe-forge.org 
Enviadas: Sexta-feira, 12 de Abril de 2013 11:14:42 
Assunto: Re: [Pw_forum] neb-GPU 


Dear Eugenio, 


please subscribe the specific QE-GPU list here: 
http://qe-forge.org/mailman/listinfo/q-e-gpgpu ). I do not regularly look at 
the pw_forum ML so it is much better for all users to address their issues and 
questions to the other specific mailing-list. 


Regarding your problem, I am assuming you are running on a workstation because 
despite you enabled both parallel and OpenMP you run your test using only 1 MPI 
process and 4 OpenMP threads. Ehm... wrong. If you are using a workstation with 
a single GPu there is no point to user --enable-parallel (and use one single 
MPi process). Please recompile your code disabling parallel 
(--disable-parallel) and run it in serial. When you fix this, if the problem 
persists try to disable MAGMA (--disable-magma). 


An observation about your specific problem: you are running a NEB calculation 
on 2 images using 3 hydrogen atoms. This system is really small, I do not think 
you are going to exploit any interesting GPU acceleration running it. 


One thing is not from you email: whatis the model of the GPU you are using? Is 
20 the correct compute capability? 


HTH 


Regards, 
Filippo 







-- 
Mr. Filippo SPIGA, M.Sc. 
http://filippospiga.me ~ skype: filippo.spiga 


?Nobody will drive us out of Cantor's paradise.? ~ David Hilbert 




On Thu, Apr 11, 2013 at 6:03 PM, Eugenio Furtado < eugeniofs at iq.ufrj.br > 
wrote: 




...input and output (now really attached) 


Eugenio 

-- 
Eugenio Furtado de Souza 
Laboratorio de Modelagem Molecular-LABMMOL 
Universidade Federal do Rio de Janeiro 
Av. Athos da Silveira Ramos No 149, CT, Bloco A, sala 609 
tel: (21) 2562-7132 
Cidade Universit?ria, Ilha do Fund?o, Rio de Janeiro-RJ, CEP 21941-909 

"Together the parts makes a giant..." 

_______________________________________________ 
Pw_forum mailing list 
Pw_forum at pwscf.org 
http://pwscf.org/mailman/listinfo/pw_forum 


_______________________________________________ 
Pw_forum mailing list 
Pw_forum at pwscf.org 
http://pwscf.org/mailman/listinfo/pw_forum 

-- 
Eugenio Furtado de Souza 
Laboratorio de Modelagem Molecular-LABMMOL 
Universidade Federal do Rio de Janeiro 
Av. Athos da Silveira Ramos No 149, CT, Bloco A, sala 609 
tel: (21) 2562-7132 
Cidade Universit?ria, Ilha do Fund?o, Rio de Janeiro-RJ, CEP 21941-909 

"Together the parts makes a giant..." 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: 
http://pwscf.org/pipermail/pw_forum/attachments/20130412/5c8b51a8/attachment.html
 

Reply via email to