Skip to Content.
Sympa Menu

forum - Re: [abinit-forum] How to use mkmem?

forum@abinit.org

Subject: The ABINIT Users Mailing List ( CLOSED )

List archive

Re: [abinit-forum] How to use mkmem?


Chronological Thread 
  • From: "张�s" <zhangting1980323@gmail.com>
  • To: forum@abinit.org
  • Subject: Re: [abinit-forum] How to use mkmem?
  • Date: Mon, 19 Feb 2007 20:07:40 +0800
  • Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=beta; h=received:message-id:date:from:to:subject:in-reply-to:mime-version:content-type:references; b=jGfGxt6nJ7lDBycEXd874Wa7P7YZlKApL9i3wvTz669+7LZyVjj61RUrpfwS2vm5GmUf4zOAn4jZFTNJyMJbpTAEjFUDgWT31Du8qrDKLe9uv/xOigA2hUO/2bAc8HOZog5XopA9wqPBdwy0Nwg+VfCf/jYUIJGdndBxj5p0x5k=

Dear Matthieu Verstraete:    

    
Thanks very much for your kindly reply. I'll explain my question in detail here. My abinit version is 4.6.5, and I've compiled it on two platforms, one is a 4 CPU Opteron machine, with 8Gb memory, and the other is a cluster, with 4 nodes, each node has two 3.06Ghz Xeon CPUs, and 2Gb's memory. The cluster totally has 8 CPUs and 8Gb's memory. My mpi version is MPICH-1.2.7, compiler is intel fortran compiler 9.0.021. The available disk space mounted under /home are about 18Gb in the Opteron machine and 6Gb in the head node of the cluster(The head node's disk is mounted during NFS to other nodes), so I think the available space in disk is enough. Scratch space are 8Gb and 2Gb per node,  same as their memory, respectively. The problems, are same at these two different platforms.
    Reading the output file with mkmem /= 0 and with just one CPU, It says the job will need 2356Mb mem. With top command, I saw a 2287Ms'  mem usage  before the job crashes with a  "Segmentation fault" message. The swap maximum useage is about 1G. The job failure happens just before the first 2DEtotal is carried out, in 1st-order perturbative wavefucntion's calculation.
    Another symptom, when I use mpirun command  to run a job with mkmem=0, it will fail as mentioned above(the error message in log file is:
ERROR -
-P-0000   Reading option of rwwf. Trying to read
-P-0000   the (npw,nspinor,nband) record of a wf file, unit=   3
-P-0000   gave iostat=  -1. Your file is likely not correct.
-P-0000   Action: check your input wf file. ),
but, when I directly run a single CPU job without command "mpirun", the job accomplishes perfectly. Maybe this can due to a file permission problem?  Thanks again for your patience!


Regards
    
                                      Zhang Ting
                                      Peking Univ
                                      Feb, 19th, 2007
                                                                


2007/2/19, Matthieu Verstraete < mjv500@york.ac.uk>:

Hello,

for any crashes of this type it is even more imperative you follow the
nettiquette. You need to tell us which version of abinit you're using,
which platform, with how much memory, etc... otherwise it's quite
pointless. Could also be a lack of disk space on scratch or home
(wherever the mkmem wavefunctions are being kept).

Superficially your CNT input file looks ok. Monitor the abinit memory with
top to see if it does go above 2 or 4 gb in the mkmem /= case. May not be
that at all.

Matthieu

--
================================================================
Dr. Matthieu Verstraete                 mailto:mjv500@york.ac.uk
Dept. of Physics, University of York,     tel: +44 1904 43 22 08
Heslington, YO10 5DD York, United Kingdom fax: +44 1904 43 22 14




Archive powered by MHonArc 2.6.16.

Top of Page