Re: [gmx-users] checkpointing in 4.0.7
Hi Mark, thanks a lot for your reply. I checked the 4.0.7 version installed on blugene and it also doesnt write checkpoints during simulations. How it was compiled by support team i dont know, but in my compilation i mostly used options from gromacs wiki page. I think i'll not trace this issue now (due to time lack), but use tpbconv for continuation runs. I though that someone perhaps knows already solution. Best regards, Leonid Mark Abraham wrote: On 25/03/2010 10:01 PM, Dr. Leonid Yelash wrote: Hello everybody, does anyone have a similar problem writing checkpoints from Gromacs 4.0.7? I have two Gromacs versions, 4.0.5 and 4.0.7, compiled for bluegene. The 4.0.5 writes state.cpt during mdrun and at the end of simulation too. The 4.0.7 does it at the end of the run only, but not during the simulation. All input files (grompp.mdp, topol.top, conf) used to start mdrun are same. To set/change the -cpt option of mdrun didnt help. Will appreciate for any hint ! I'm not aware of any BlueGene-related issue here. It's possible something is causing the I/O buffering to work differently. You could test this by starting a sufficiently long 4.0.7 simulation and seeing what happens when you kill its job in LoadLeveler. Were the two compilations done the same way linking to the same libraries? Also, do see if anything here http://www.gromacs.org/About_Gromacs/Release_Notes/Revisions_in_4.0 pertains to your situation. Mark -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users Please search the archive at http://www.gromacs.org/search before posting! Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. Can't post? Read http://www.gromacs.org/mailing_lists/users.php
[gmx-users] checkpointing in 4.0.7
Hello everybody, does anyone have a similar problem writing checkpoints from Gromacs 4.0.7? I have two Gromacs versions, 4.0.5 and 4.0.7, compiled for bluegene. The 4.0.5 writes state.cpt during mdrun and at the end of simulation too. The 4.0.7 does it at the end of the run only, but not during the simulation. All input files (grompp.mdp, topol.top, conf) used to start mdrun are same. To set/change the -cpt option of mdrun didnt help. Will appreciate for any hint ! Best regards, Leonid -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users Please search the archive at http://www.gromacs.org/search before posting! Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. Can't post? Read http://www.gromacs.org/mailing_lists/users.php
Re: [gmx-users] msd
Hello together, i refer here to the problem with msd mentioned at the beginning of September (see also below): i calculate msd this time for monomers (not for molecules as before) and i still observe finite size effects for chains if they include fixed bonds ([constrains]), however this effect disappears when i replace fixed bonds by lj+fene bonds. Is it possible that lincs changes (slightly) the velocity field and hence the dynamics? Regards, Leonid Berk Hess wrote: Hi, g_msd calculates the msd for molecules, not for atoms. I guess that would explain the result when you half the chain length. It might also explain the box size effects, since whole chains will still have a reduction in MSD due to periodiciy with a box of 3 Rg. If you run g_msd without -mol you get the MSD per atom. Berk From: yel...@uni-mainz.de To: gmx-users@gromacs.org Date: Thu, 3 Sep 2009 15:22:40 +0200 Subject: [gmx-users] msd Hello together, i have a problem with msd. i simulate a melt of chain molecules consisting of 116 united atoms per chain. then i copy the box n times in x,y,z directions, repeat simulations and get a slightly different msd curves: in the time range between 1 ps and 10 ps it flattens out decreasing the curvature for larger boxes and at later time (10ps) follows at higher msd just parallel to the curve for smaller box. the boxe sizes were tried between 3 to 15 Rg's (gyration radii), so if it would be the finite size effect it should already disappear, but it doesnt. I also tried with different thermostats (sd, md with nose-hoover), different tau_t and also switching off the removal of com (then the box starts to fly after awhile) To check from the other side if it could be the finite size effect i reduce the length of the chains from 116 to 58 and the whole msd curve shifted up, also in ballistic regime where one can expect no influence of chain configuration. interesting that the shift is by factor of ~2 (although is not exactly): perhaps i'm just doing sth wrong! msd was calculated for chains using: g_msd -mol diff_mol.xvg is anybody aware of such things or has ideas what was done wrong? Thanks a lot in advance for any suggestions! Regards, Leonid ___ gmx-users mailing list gmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users Please search the archive at http://www.gromacs.org/search before posting! Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. Can't post? Read http://www.gromacs.org/mailing_lists/users.php Express yourself instantly with MSN Messenger! MSN Messenger http://clk.atdmt.com/AVE/go/onm00200471ave/direct/01/ ___ gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users Please search the archive at http://www.gromacs.org/search before posting! Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. Can't post? Read http://www.gromacs.org/mailing_lists/users.php
[gmx-users] -append on jugene overwrites trr
Hello all together, in my compilation of Gromacs 4.0.5 on JUGENE (a Blue Gene/P machine in Jülich), the -append option does not work properly. When i restart a simulation from a checkpoint, the mdrun recognizes the checkpoint and -append (one can see it in md.log), but instead of appending to md.log, traj.trr, ener.edr, it overwrites them. The file offsets in state.cpt (printed by gmxdump -cp) are shown here: number of output files = 3 output filename = md.log file_offset_high = 0 file_offset_low = 21265 output filename = traj.trr file_offset_high = 0 file_offset_low = 9021240 output filename = ener.edr file_offset_high = 0 file_offset_low = 5872 The same distribution compiled for JUMP (a Power6 machine running under AIX) works correctly. Details to the machines can be found here http://www.fz-juelich.de/jsc/service/sco_ibmBGP http://www.fz-juelich.de/jsc/service/sco_ibmP6 Does anybody know how to solve/work around? At the moment i have to store continuations of a simulation run into different directories, which complicates analysing data, e.g. the calculation of time-dependent properties for the whole run. Many thanks in advance for any hints! Leonid ___ gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users Please search the archive at http://www.gromacs.org/search before posting! Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. Can't post? Read http://www.gromacs.org/mailing_lists/users.php
