[Pw_forum] Insufficient Virtual Memory

[Axel Kohlmeyer]

On Thu, 26 Feb 2009, Vo, Trinh wrote:

TV> Dear Axel,
TV> 

TV> Yes, I read the README file.  I first just run "run_example" to get 
TV> the input files (I wanted to see what they are).  I then run each 
TV> job individually.  Since I saw in one of previous emails, you 
TV> mentioned the following link, which is related to the benchmark of 
TV> the 8cores:

TV> http://cp2k.googlegroups.com/web/cp2k-water-bench-cmm.png?gda=cZtL20oAAABKcjB2R2-kyOTBNGr0GXCNfKm1opB3ppoZQi3FVKdigwDOq5Mi_ip_cCKVysjiE_BA1utot6QxKJxvVe1Mw7MU_e3Wg0GnqfdKOwDqUih1tA

TV> In the plot, I saw the result for 512 H2O and 4core/nodes.  Thus, I 
TV> tried to repeat all the calculations to compare, but I could not run 
TV> that case.

this graph was done with a _very_ different code, cp2k, that
uses a method (quickstep) different from the CP in cp.x and has 
therefore different memory requirements and system size scaling.
the README of example21 explicitly warns you about running out 
of memory. give the current difference in speed between hard 
drives and main memory, every calculation where you run into 
using swap is not worth the effort (unlike 10 years ago).

just for reference: because of the difference in methods cp.x 
should scale better with the number of processors, whereas 
cp2k should scale better with the problem set size than cp.x
with the same number of processors.

that being said, cp2k demonstrates the problem of overloading 
memory and communication channels very nicely and thus the
graphs are a nice point of reference to showcase how bad it
can get and that the not so obvious choice of not using all
cpus actually gives you the better performance. 

cheers,
   axel.

TV> 
TV> Thank you,
TV> 
TV> Trinh    
TV> 
TV> 
TV> 
TV> -----Original Message-----
TV> From: pw_forum-bounces at pwscf.org [mailto:pw_forum-bounces at pwscf.org] On Behalf Of Axel Kohlmeyer
TV> Sent: Thursday, February 26, 2009 1:04 PM
TV> To: PWSCF Forum
TV> Subject: Re: [Pw_forum] Insufficient Virtual Memory
TV> 
TV> On Thu, 26 Feb 2009, Vo, Trinh wrote:
TV> 
TV> TV> Dear PWSCF users,
TV> TV> 
TV> TV> I was running a benchmark (using example21 in QE) for my computer cluster (total 18 nodes) to compare the results with those posted on web link that was mentioned by Axel in some previous forum emails..
TV> TV> 
TV> TV> When I performed the calculation for the case of 512 H2O with 4cores/node, the job crashed with the error:
TV> TV> 
TV> TV>  "forrtl: severe (41): insufficient virtual memory".
TV> TV> 
TV> 
TV> TV> I do not know how to deal with this.  Any help or suggestion is 
TV> TV> appreciated.  For other cases, the runs were fine.
TV> 
TV> have you _read_ the README file in that directory???
TV> 
TV> axel.
TV> 
TV> 
TV> TV> 
TV> TV> Below are the specifications of my cluster:
TV> TV> 
TV> TV> Intel CPU Xeon X5450 Quad Core 3.0GHz with 12M cache 133MHz FSB
TV> TV> DDRII667 4GB RAM FBD ECC (D-667D2D4F5/4GB)
TV> TV> 
TV> TV> For compilation, I used the following:
TV> TV> QE ver. 4.0.3
TV> TV> ifort 10.1015, MKL  10.0.3.020, impi 3.1
TV> TV> internal fftw
TV> TV> 
TV> TV> I also attach here the output file.
TV> TV> 
TV> TV> 
TV> TV> Thank you,
TV> TV> 
TV> TV> Trinh Vo
TV> TV> 
TV> TV> 
TV> TV> 
TV> TV> 
TV> TV> 
TV> 
TV> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.