The ABINIT Users Mailing List ( CLOSED )

[Paul Fons <paul-fons@aist.go.jp>]

The problem turned out to be simple -- usually abinit has fairly robust error checking on input parameters, but apparently not for parallel processing variables (yet).  The number of parallel process I was invoking did not add up to the number of cores and it lead to a (misleading) crash.  Entering a set of parameters that matched to available cores fixed the problem and the input file runs just fine although there is a fair amount of overhead over the gigabit ethernet connection between the different cpus.  

Paul

On Sep 25, 2009, at 2:52 PM, St Druid wrote:

Hi, Paul:

 In my view, MAC OS 10 Leopard is  a Desktop Operation System, so that no one can full utilize the system resource, whatever how much processes are running, I have tried to compile PWSCF in Leopard, but I can't obtain a near 100% CPU utilization. I think it is optimized by Apple. co. in order to enhance the quantity in daily computation.

So, i just wondered that have you full utilized the CPU resources?.

Best regards.

On Thu, Sep 24, 2009 at 4:04 PM, BOTTIN Francois <francois.bottin@cea.fr> wrote:

Dear Paul,

I feel it's coming from iprcel 45.
Your job seems to crash within the susceptibility matrix calculation.
If this part is allowed using paral_kgb=1, at this time it is very time consuming and ressource demanding.
So it is only allowed for testing purpose.

If not, please sent your whole ouptput and log files.

Regards,
Francois

--
B.W. Dong
State Key Laboratory of Superhard Materials,
Jilin University, Changchun, China. 130012

tq02ksu@gmail.com [ dbw07@mails.jlu.edu.cn ]