PC GAMESS/Firefly-related discussion club



Learn how to ask questions correctly


Re: Running pcgamess on skif-mgu supercluster

Alex Granovsky
gran@classic.chem.msu.su


Hi,

as far as I remember that really strange cluster has only 4 GB per eight-core box, and there is no swap partition at all; and you are trying to allocate 16 GB...

Regards,
Alex Granovsky


On Sun Mar 8 '09 5:55pm, DimitryASuplatov wrote
-----------------------------------------------
>Hello,

>I am trying to run pcgamess on skif supercomputer of MSU.
>I use the following distribution "Direct PC GAMESS/Firefly version 7.1.F download links, Linux MVAPICH/libibverbs, dynamically linked NPTL-based version" since its the only one MVAPICH based distribution available for download.

>I run firefly using the following options:
> $CONTRL SCFTYP=RHF RUNTYP=OPTIMIZE NPRINT=7 MULT=1 ICHARG=-1
>      MAXIT=1000 COORD=UNIQUE  LOCAL=NONE  $END
> $SYSTEM MWORDS=256 TIMLIM=2700 $END
> $P2P P2P=.TRUE. XDLB=.TRUE. $END
> $STATPT NSTEP=200 OPTTOL=0.0001 $END
> $STATPT  METHOD=GDIIS HESS=CALC $END
> $BASIS GBASIS=N31 NGAUSS=6 NDFUNC=0 NFFUNC=0 NPFUNC=0
>     DIFFSP=.FALSE. DIFFS=.FALSE. $END
> $GUESS GUESS=HUCKEL $END
> $DATA
>...
> $END

>Firefly starts without complaints on 64 nodes. Since I use non-direct method it creates 64 folders to store the data.
>The problem is that they are all almost empty - 3.7MB constantly per folder. I have no output to PUNCH file (!) and my .out file contains only first input cards echo.
>Nothing seems to change during 24 hours run. In the end when duration exceeds time limit jobs hangs without any output.

>What is wrong with my job? Any suggestions?
>Thanks. I appreciate your time.

>SDA

[ This message was edited on Mon Mar 9 '09 at 9:36pm by the author ]


[ Previous ] [ Next ] [ Index ]           Mon Mar 9 '09 9:36pm
[ Reply ] [ Edit ] [ Delete ]           This message read 1181 times