[gmx-developers] Understanding disparity in field values of "t_state" class

[Erik Lindahl]

If you are on release-2018 I think you have the PaddedRVecVectors, and then
it is handled by hiding the allocation size inside the class.

Note that the exact size of the *allocation* is not relevant to determine a
unique physical *state* for your system. The extra positions are not used,
it's merely an optimization to make the memory behave better. The physics
will be the same even if we restore the state to another structure without,
or with different, padding.

Cheers,

Erik

On Thu, Mar 1, 2018 at 10:04 PM, Sanyam Kapoor <sanyam at nyu.edu> wrote:

> @Mark Sorry to have left the version information out. I'm currently pinned
> to version *v2018*.
>
> @Erik I can't seem to find another count variable. Do you mind pointing me
> to it? Also, just to put it out, I'm trying to inspect this after the
> threads have been joined (right before the return). Is there a way for me
> to find one-to-one correspondence between the atom number and this state
> structure?
>
> Thank you so much for the prompt responses!
>
>
> On Thu, Mar 1, 2018 at 3:59 PM Mark Abraham <mark.j.abraham at gmail.com>
> wrote:
>
>> Hi,
>>
>> What Erik said, plus also this behaviour varies whether there is more
>> than one rank, whether there are virtual sites, and with GROMACS version.
>>
>> Mark
>>
>> On Thu, Mar 1, 2018 at 9:57 PM Erik Lindahl <erik.lindahl at gmail.com>
>> wrote:
>>
>>> Hi Sanyam,
>>>
>>> Not entirely sure what your debugger reports, but it is likely due
>>> either to padding (which we do to avoid false sharing between threads), or
>>> over-allocation (which we do to avoid repeatedly re-allocating and freeing
>>> memory when particles drift between nodes during domain decomposition).
>>>
>>> This is also why we have separate fields to keep track of the number of
>>> atoms vs. the number of allocated entries.
>>>
>>> Cheers,
>>>
>>> Erik
>>>
>>> On Thu, Mar 1, 2018 at 9:52 PM, Sanyam Kapoor <sanyam at nyu.edu> wrote:
>>>
>>>> Hello everyone,
>>>>
>>>> I was trying to understand fields in the t_state class and it looks
>>>> like there is some disparity during the run which I cannot make sense of. I
>>>> ran the code for two topologies experiments with the following arguments
>>>>
>>>> mdrun -s m.tpr -rerun m.gro -e m.edr -g m.log
>>>>
>>>> 1. A topology which had *1960 atoms *but the *fields for coordinates
>>>> (x) and velocities (v) had 1968* items.
>>>> 2. A topology which had *1793 atoms *but the *fields for coordinates
>>>> (x) and velocities (v) had 1808* items.
>>>>
>>>> These numbers that I am reporting have been retrieved while debugging
>>>> for the fields natoms, x and v of t_state.
>>>>
>>>> The comments for t_state state that the lists should have been of the
>>>> same size as number of atoms. However this doesn't seem to be the case.
>>>> Could somebody verify if this is expected? If yes, why so?
>>>>
>>>> Thank you!
>>>> --
>>>> Regards,
>>>> Sanyam Kapoor
>>>> Masters in Computer Science
>>>> Courant Institute, New York University
>>>> https://www.sanyamkapoor.com
>>>>
>>>> --
>>>> Gromacs Developers mailing list
>>>>
>>>> * Please search the archive at http://www.gromacs.org/
>>>> Support/Mailing_Lists/GMX-developers_List before posting!
>>>>
>>>> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>>>
>>>> * For (un)subscribe requests visit
>>>> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers
>>>> or send a mail to gmx-developers-request at gromacs.org.
>>>>
>>>
>>>
>>>
>>> --
>>> Erik Lindahl <erik.lindahl at dbb.su.se>
>>> Professor of Biophysics, Dept. Biochemistry & Biophysics, Stockholm
>>> University
>>> Science for Life Laboratory, Box 1031, 17121 Solna, Sweden
>>> --
>>> Gromacs Developers mailing list
>>>
>>> * Please search the archive at http://www.gromacs.org/
>>> Support/Mailing_Lists/GMX-developers_List before posting!
>>>
>>> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>>
>>> * For (un)subscribe requests visit
>>> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers
>>> or send a mail to gmx-developers-request at gromacs.org.
>>
>> --
>> Gromacs Developers mailing list
>>
>> * Please search the archive at http://www.gromacs.org/
>> Support/Mailing_Lists/GMX-developers_List before posting!
>>
>> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>
>> * For (un)subscribe requests visit
>> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers
>> or send a mail to gmx-developers-request at gromacs.org.
>
> --
> Regards,
> Sanyam Kapoor
> Masters in Computer Science
> Courant Institute, New York University
> https://www.sanyamkapoor.com
>
> --
> Gromacs Developers mailing list
>
> * Please search the archive at http://www.gromacs.org/
> Support/Mailing_Lists/GMX-developers_List before posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers
> or send a mail to gmx-developers-request at gromacs.org.
>



-- 
Erik Lindahl <erik.lindahl at dbb.su.se>
Professor of Biophysics, Dept. Biochemistry & Biophysics, Stockholm
University
Science for Life Laboratory, Box 1031, 17121 Solna, Sweden
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20180301/bfbe1398/attachment.html>