[Pw_forum] why my pw.x run with low efficiency?

[Lorenzo Paulatto]

On Dom, Settembre 21, 2008 11:44, vega wrote:
>  So could you estimate
>  that how many memory
>  will be good enough for 40 [TiO2] units( 120 atoms)?

No way, it depends strongly on cutoff and cell size but weakly on number
and types of atoms. Furthermore, I can't do it even with the correct data
avilable: if I could I would put it in the code!

I'm not really convinced that you have a swap problem, truly your
performance are degraded, but not as much as I would expect from swap.

You should ask the sys administrator to disable swap usage, or you can use
the following hack to prevent the code from using swap space. Before
running the pw code you can set the max memory limit per process equal to
the amount of free RAM:
FREE_RAM=$(free|grep buffers/cache|awk '{print $4}')
ulimit -v $FREE_RAM

In this way if the program will try to use more RAM than available it will
be killed, and you'll surely notice.
Read the ``ulimit'' help and ``free '' manpage for details on what you are
doing.

regards

-- 
Lorenzo Paulatto
SISSA  &  DEMOCRITOS (Trieste)
phone: +39 040 3787 511
skype: paulatz
www:   http://people.sissa.it/~paulatto/



----------------------------------------------------------------
  SISSA Webmail https://webmail.sissa.it/
  Powered by SquirrelMail http://www.squirrelmail.org/