[Pw_forum] 5.0.1 "cannot allocate spsi" in cegterg

[Filippo Spiga]

Dear Ahtoh,

On Jan 19, 2013, at 1:44 PM, Антон Литвиненко <tenebrosus.scriptor at gmail.com> wrote:
>  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
>      Error in routine  cegterg (5014):
>       cannot allocate spsi
>  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

This can happen if PWscf tries to allocate more memory than available on the system. Possible solutions depend by how you run the code….


> The problem persists when turning on/off OpenMP and CUDA; both with GNU (4.4.6) and Intel (13.0.1) compilers, MKL and reference BLAS.
> 
> System: Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz; 4 cores; 16 Gb of RAM; 32-bit CentOS 6.2 (PAE kernel 2.6.32-220.13.1.el6.i686).

and if I correctly guess you are running the serial calculation using OpenMP. Since you running with lot K point I can say that before the ortho-diagonalization step due to some additional allocations the required memory blows up. I believe that if you can share also the upper part of the output is much better to understand your problem because in a specific point QE prints the memory required for some data structures allocated during the execution. 

Btw, are you trying also QE-GPU?

Cheers,
Filippo


(please mention name and institution if it is possible ;-)

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

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

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20130119/9667b185/attachment.html>