On Wed, Jun 1, 2011 at 6:11 AM, Lorenzo Gontrani <l.gontrani.caspur.it>wrote:
> Dear all, here it is the output of a ptraj dihedral analysis on the same
> system, for both ASCII and BINARY trajectory. Starting from frame 14,
> the values start to diverge significantly.
>
I don't understand what you're doing here. Is the ASCII trajectory the
*same* as the NetCDF trajectory? Were they from 2 different simulations?
Or were they from the same simulation? If they were from the same
simulation, what was the original format (NetCDF or ASCII)?
That said, Binary trajectories have much higher precision, so the actual
values of the coordinates are in fact different between binary and ASCII
(since ASCII truncates to 3 decimal places). Inserting this loss of
precision into a dihedral calculation could lead to the results you're
seeing here.
If they're actually 2 separate simulations, then there's no reason to expect
them to be identical, as machine-precision limits reproducibility,
*especially* across non-identical hardware. In fact, probably the *only*
way to ensure reproducibility is to use the GPU code, which is completely
deterministic.
HTH,
Jason
> BINARY ASCII
>
> 1.00 -147.958668 | 1.00 -147.942612
> 2.00 -72.048594 | 2.00 -72.067980
> 3.00 18.534202 | 3.00 18.516501
> 4.00 82.936517 | 4.00 82.915310
> 5.00 141.777751 | 5.00 141.796965
> 6.00 26.969903 | 6.00 26.895879
> 7.00 -67.860564 | 7.00 -67.880553
> 8.00 -130.117973 | 8.00 -130.058275
> 9.00 -83.235696 | 9.00 -83.207713
> 10.00 -44.748722 | 10.00 -44.763429
> 11.00 -134.239920 | 11.00 -134.211415
> 12.00 -135.534592 | 12.00 -135.964199
> 13.00 -120.563889 | 13.00 -120.421923
> 14.00 148.857056 | 14.00 155.068374
> 15.00 145.166697 | 15.00 179.217652
> 16.00 123.163784 | 16.00 -127.158826
> 17.00 163.278049 | 17.00 -127.651488
> 18.00 169.775862 | 18.00 -116.616523
> 19.00 170.453577 | 19.00 -118.063325
> 20.00 -168.840722 | 20.00 -146.686249
>
> I hope that someone can help someway..
>
> Lorenzo
>
> --
> =======================================
> Lorenzo Gontrani
> Research associate of EDXD group
> University of Rome "La Sapienza"
>
> GSM +39 338 7615798
> WORK +39 06 49693273
> Email l DOT gontrani AT caspur DOT it
> Webpage: http://webcaminiti/gontrani.html
> =======================================
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jun 01 2011 - 06:30:02 PDT
