<html dir="ltr">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
<meta name="GENERATOR" content="MSHTML 8.00.6001.19190">
<style id="owaParaStyle">P {
MARGIN-TOP: 0px; MARGIN-BOTTOM: 0px
}
</style>
</head>
<body fPStyle="1" ocsi="0">
<div style="direction: ltr;font-family: Tahoma;color: #000000;font-size: 10pt;">
<div style="FONT-FAMILY: Tahoma; DIRECTION: ltr; COLOR: rgb(0,0,0); FONT-SIZE: 10pt">
Hi,<br>
<br>
I have Gromacs-GPU<a></a> version 4.5.5 and GTX<a></a> 580. <br>
I ran dhfr-solv-PME<a></a> benchmark test (see below) and my run was killed after couple of hours exhausting all the computer memory, including the swap (2G<a></a> + 4G<a></a> swap).<br>
Has anyone encountered this problem? What is wrong? <br>
<br>
Thanks, Efrat<a></a></div>
<div style="FONT-FAMILY: Tahoma; DIRECTION: ltr; COLOR: rgb(0,0,0); FONT-SIZE: 10pt">
</div>
<div style="FONT-FAMILY: Tahoma; DIRECTION: ltr; COLOR: rgb(0,0,0); FONT-SIZE: 10pt">
> mdrun-gpu<a></a> -device "OpenMM:platform<a></a>=Cuda,memtest<a></a>=15,deviceid<a></a>=0,force<a></a>-device=yes" -deffnm<a></a> md<a></a><br>
:-) G R O M A C S (-:<br>
<br>
Great Red Oystrich<a></a> Makes All Chemists Sane<br>
<br>
:-) VERSION 4.5.5 (-:<br>
<br>
Written by Emile Apol<a></a>, Rossen<a></a> Apostolov<a></a>, Herman J.C. Berendsen<a></a>,<br>
Aldert<a></a> van Buuren<a></a>, Pär<a></a> Bjelkmar<a></a>, Rudi van Drunen<a></a>, Anton Feenstra<a></a>, <br>
Gerrit<a></a> Groenhof<a></a>, Peter Kasson, Per Larsson, Pieter Meulenhoff<a></a>, <br>
Teemu<a></a> Murtola<a></a>, Szilard Pall, Sander Pronk<a></a>, Roland Schulz,
<br>
Michael Shirts, Alfons<a></a> Sijbers<a></a>, Peter Tieleman<a></a>,<br>
<br>
Berk<a></a> Hess, David van der<a></a> Spoel<a></a>, and Erik Lindahl<a></a>.<br>
<br>
Copyright (c) 1991-2000, University of Groningen, The Netherlands.<br>
Copyright (c) 2001-2010, The GROMACS<a></a> development team at<br>
Uppsala University & The Royal Institute of Technology, Sweden.<br>
check out http://www.gromacs.org for more information.<br>
<br>
This program is free software; you can redistribute it and/or<br>
modify it under the terms of the GNU General Public License<br>
as published by the Free Software Foundation; either version 2<br>
of the License, or (at your option) any later version.<br>
<br>
:-) mdrun-gpu<a></a> (-:<br>
<br>
Option Filename Type Description<br>
------------------------------------------------------------<br>
-s md.tpr<a></a> Input Run input file: tpr<a></a> tpb<a></a> tpa<a></a><br>
-o md.trr<a></a> Output Full precision trajectory: trr<a></a> trj<a></a> cpt<a></a><br>
-x md.xtc<a></a> Output, Opt. Compressed trajectory (portable xdr<a></a> format)<br>
-cpi<a></a> md.cpt<a></a> Input, Opt. Checkpoint file<br>
-cpo<a></a> md.cpt<a></a> Output, Opt. Checkpoint file<br>
-c md.gro<a></a> Output Structure file: gro<a></a> g96<a></a> pdb<a></a> etc.<br>
-e md.edr<a></a> Output Energy file<br>
-g md.log<a></a> Output Log file<br>
-dhdl<a></a> md.xvg<a></a> Output, Opt. xvgr<a></a>/xmgr<a></a> file<br>
-field md.xvg<a></a> Output, Opt. xvgr<a></a>/xmgr<a></a> file<br>
-table md.xvg<a></a> Input, Opt. xvgr<a></a>/xmgr<a></a> file<br>
-tablep<a></a> md.xvg<a></a> Input, Opt. xvgr<a></a>/xmgr<a></a> file<br>
-tableb<a></a> md.xvg<a></a> Input, Opt. xvgr<a></a>/xmgr<a></a> file<br>
-rerun md.xtc<a></a> Input, Opt. Trajectory: xtc<a></a> trr<a></a> trj<a></a> gro<a></a> g96<a></a> pdb<a></a> cpt<a></a><br>
-tpi<a></a> md.xvg<a></a> Output, Opt. xvgr<a></a>/xmgr<a></a> file<br>
-tpid<a></a> md.xvg<a></a> Output, Opt. xvgr<a></a>/xmgr<a></a> file<br>
-ei<a></a> md.edi<a></a> Input, Opt. ED sampling input<br>
-eo<a></a> md.edo<a></a> Output, Opt. ED sampling output<br>
-j md.gct<a></a> Input, Opt. General coupling stuff<br>
-jo<a></a> md.gct<a></a> Output, Opt. General coupling stuff<br>
-ffout<a></a> md.xvg<a></a> Output, Opt. xvgr<a></a>/xmgr<a></a> file<br>
-devout md.xvg<a></a> Output, Opt. xvgr<a></a>/xmgr<a></a> file<br>
-runav<a></a> md.xvg<a></a> Output, Opt. xvgr<a></a>/xmgr<a></a> file<br>
-px<a></a> md.xvg<a></a> Output, Opt. xvgr<a></a>/xmgr<a></a> file<br>
-pf<a></a> md.xvg<a></a> Output, Opt. xvgr<a></a>/xmgr<a></a> file<br>
-mtx<a></a> md.mtx<a></a> Output, Opt. Hessian matrix<br>
-dn<a></a> md.ndx<a></a> Output, Opt. Index file<br>
-multidir<a></a> md<a></a> Input, Opt., Mult<a></a>. Run directory<br>
<br>
Option Type Value Description<br>
------------------------------------------------------<br>
-[no]h bool<a></a> no Print help info and quit<br>
-[no]version bool<a></a> no Print version info and quit<br>
-nice int<a></a> 0 Set the nicelevel<a></a><br>
-deffnm<a></a> string md<a></a> Set the default filename for all file options<br>
-xvg<a></a> enum<a></a> xmgrace<a></a> xvg<a></a> plot formatting: xmgrace<a></a>, xmgr<a></a> or none<br>
-[no]pd bool<a></a> no Use particle decompostion<a></a><br>
-dd<a></a> vector 0 0 0 Domain decomposition grid, 0 is optimize<br>
-npme<a></a> int<a></a> -1 Number of separate nodes to be used for PME<a></a>, -1<br>
is guess<br>
-ddorder<a></a> enum<a></a> interleave DD node order: interleave, pp_pme<a></a> or cartesian<a></a><br>
-[no]ddcheck<a></a> bool<a></a> yes Check for all bonded interactions with DD<br>
-rdd<a></a> real 0 The maximum distance for bonded interactions with<br>
DD (nm), 0 is determine from initial coordinates<br>
-rcon<a></a> real 0 Maximum distance for P-LINCS<a></a> (nm), 0 is estimate<br>
-dlb<a></a> enum<a></a> auto Dynamic load balancing (with DD): auto, no or yes<br>
-dds<a></a> real 0.8 Minimum allowed dlb<a></a> scaling of the DD cell size<br>
-gcom<a></a> int<a></a> -1 Global communication frequency<br>
-[no]v bool<a></a> no Be loud and noisy<br>
-[no]compact bool<a></a> yes Write a compact log file<br>
-[no]seppot<a></a> bool<a></a> no Write separate V and dVdl<a></a> terms for each<br>
interaction type and node to the log file(s)<br>
-pforce<a></a> real -1 Print all forces larger than this (kJ/mol nm)<br>
-[no]reprod<a></a> bool<a></a> no Try to avoid optimizations that affect binary<br>
reproducibility<br>
-cpt<a></a> real 15 Checkpoint interval (minutes)<br>
-[no]cpnum<a></a> bool<a></a> no Keep and number checkpoint files<br>
-[no]append bool<a></a> yes Append to previous output files when continuing<br>
from checkpoint instead of adding the simulation<br>
part number to all file names<br>
-maxh<a></a> real -1 Terminate after 0.99 times this time (hours)<br>
-multi int<a></a> 0 Do multiple simulations in parallel<br>
-replex<a></a> int<a></a> 0 Attempt replica exchange every # steps<br>
-reseed int<a></a> -1 Seed for replica exchange, -1 is generate a seed<br>
-[no]ionize bool<a></a> no Do a simulation including the effect of an X-Ray<br>
bombardment on your system<br>
-device string OpenMM:platform<a></a>=Cuda,memtest<a></a>=15,deviceid<a></a>=0,force<a></a>-device=yes Device option<br>
string<br>
<br>
<br>
Back Off! I just backed up md.log<a></a> to ./#md.log.1<a></a>#<br>
Reading file md.tpr<a></a>, VERSION 4.5.5 (double precision)<br>
<br>
Back Off! I just backed up md.edr<a></a> to ./#md.edr.1<a></a>#<br>
<br>
WARNING: OpenMM<a></a> does not support leap-frog, will use velocity-verlet<a></a> integrator.<br>
<br>
<br>
WARNING: OpenMM<a></a> supports only Andersen thermostat with the md<a></a>/md<a></a>-vv/md<a></a>-vv-avek<a></a> integrators.<br>
<br>
<br>
WARNING: OpenMM<a></a> provides contraints<a></a> as a combination of SHAKE, SETTLE and CCMA<a></a>. Accuracy is based on the SHAKE tolerance set by the "shake_tol<a></a>" option.<br>
<br>
<br>
WARNING: Non-supported GPU<a></a> selected (#0, GeForce<a></a> GTX<a></a> 580), forced continuing.Note<a></a>, that the simulation can be slow or it migth<a></a> even crash.<br>
<br>
<br>
Pre-simulation ~15s<a></a> memtest<a></a> in progress...done, no errors detected<br>
starting mdrun<a></a> 'Protein in water'<br>
-1 steps, infinite ps.<br>
Killed<br>
<br>
</div>
</div>
</body>
</html>