[gmx-developers] head mdrun broken?

Michael Shirts mrshirts at gmail.com
Thu Dec 24 17:14:51 CET 2009


> The initial temperature with integrators md and sd is exactly
> a factor of 2 too low, I guess currently the average ekin of
> a correct number and 0 is taken.

This is correct.  I didn't notice before, since the .edr temperatures
are all correct, just not the initial temperature printed to the log.
 I've put in a simple fix for now (bit of a kludge, I just multiply by
two, but it's simpler than messing with the complicated logic of
temperature calculation that needs to handle three integrators . . .

> Berendsen pcoupl works with md, but with sd the box is nan
> after the second neighbor search step.

I think this was fixed by a patch I put in last night; please let me
know if you're still getting a bug, and This bug emphasized the need
for good coding practices -- there was a one-line conditional without
brackets that I missed when I was adding velocity verlet.  If it had
brackets, I would have noticed it the first time and save a couple
hours of debugging :)

Thanks for all the error reports -- keep them coming!  Unless,
hopefully, there aren't any more!

Best,
Michael


> Michael, could you look into this?
>
> Berk
>
>> After some further checks I noticed that I now get infinite energies in a
>> simple system with the sd integrator (which didn't happen before).
>> I'll send you the run files off-list.
>>
>> Sander
>>
>>
>> On Dec 22, 2009, at 18:19 , Michael Shirts wrote:
>>
>>>> And Berendsen pressure coupling fails with a segv in pme.c - calc_idx
>>>> because of NaNs in the reciprocal box size..
>>>
>>> Berendsen seems to be working OK for me.  Sander, could you send me
>>> your .mdp so I can make sure I'm covering all the options?
>>>
>>> Obviously, my test cases before submitting are not broad enough! :)
>>>
>>> Best,
>>> Michael
>>> --
>>> gmx-developers mailing list
>>> gmx-developers at gromacs.org
>>> http://lists.gromacs.org/mailman/listinfo/gmx-developers
>>> Please don't post (un)subscribe requests to the list. Use the
>>> www interface or send it to gmx-developers-request at gromacs.org.
>>
>> --
>> gmx-developers mailing list
>> gmx-developers at gromacs.org
>> http://lists.gromacs.org/mailman/listinfo/gmx-developers
>> Please don't post (un)subscribe requests to the list. Use the
>> www interface or send it to gmx-developers-request at gromacs.org.
>>
>
>
>
> ------------------------------
>
> --
> gmx-developers mailing list
> gmx-developers at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-developers
>
>
> End of gmx-developers Digest, Vol 68, Issue 15
> **********************************************
>



More information about the gromacs.org_gmx-developers mailing list