From: Thomas Evangelidis (tevang3_at_gmail.com)
Date: Sat Dec 01 2012 - 08:39:09 CST
I have doubts that GB+aMD works correctly even on the CPU. In a personaly
communication I had with Yi Wang you developed the aMD implementation, she
said that aMD was developed for explicit solvent simulation and she is
unsure whether the dihedral boost works correctly in GB solvent simulation.
Have a look at the energies below, they were measured using the same
starting coordinates, velocities and seed. Apart from the DIHED energy,
which is higher in the dihedral aMD as expected, I can't see any similarity
in the rest of the energies. Is this normal??
cMD on one CPU:
ETITLE: TS BOND ANGLE DIHED
IMPRP ELECT VDW BOUNDARY
MISC KINETIC TOTAL TEMP
POTENTIAL TOTAL3 TEMPAVG
ENERGY: 0 558.6683 1579.5665 3885.6373
0.0000 -10480.3177 -553.1751 0.0000
0.0000 2106.1422 -2903.4785 304.4260 -5009.6206
-2846.3565 304.4260
ENERGY: 100 655.1887 1757.1695 2301.3864
0.0000 -10461.9301 -534.2638 0.0000
0.0000 2491.8642 -3790.5851 360.1790 -6282.4493
-3738.6451 382.0679
ENERGY: 200 611.7406 1662.6134 2242.0371
0.0000 -10464.3573 -533.5304 0.0000
0.0000 2284.5587 -4196.9379 330.2146 -6481.4966
-4161.1818 341.1784
dihedral boost aMD on one CPU:
ENERGY: 0 558.6683 1579.5665 3885.6373
0.0000 -10480.3177 -553.1751 0.0000
0.0000 2106.1422 -2903.4785 304.4260 -5009.6206
-2846.3565 304.4260
ENERGY: 100 602.4311 1683.1178 3224.9759
0.0000 -10571.7046 -529.4346 0.0000
0.0000 2202.8328 -3387.7816 318.4018 -5590.6144
-3352.1295 334.9764
ENERGY: 200 570.8550 1630.2132 3220.1442
0.0000 -10577.6660 -496.8820 0.0000
0.0000 2120.2538 -3533.0818 306.4657 -5653.3356
-3501.4385 315.4067
On 30 November 2012 12:26, Felipe Merino
<felipe.merino_at_mpi-muenster.mpg.de>wrote:
> Hi Jeff,
>
> Yes you are right. In the CPU it works without a problem. It is a pity
> since the GPU calculation for implicit solvent is really much faster.
>
> Regards
>
> Felipe
>
>
> On 11/29/2012 11:57 PM, Jeff Wereszczynski wrote:
>
>> Hi Felipe,
>>
>> Does it work properly with the CPU version? Or is it a problem with
>> aMD+GB in general? I believe that the only aMD that works on the GPU
>> is the dihedral acceleration (probably because those calculations are
>> done on the CPU even in the GPU NAMD).
>>
>> Jeff
>>
>> On Thu, Nov 29, 2012 at 7:39 AM, Felipe Merino <fmerinoleon_at_gmail.com>
>> wrote:
>>
>>> Dear all,
>>>
>>> I havwe been trying lately to run some accelerated molecular dynamics
>>> combined with implicit solvent using a cuda version of NAMD 2.9. However,
>>> when it comes to calculate the dV for the potential energy the
>>> electrostatics are not calculated correctly.
>>>
>>> ACCELERATED MD: STEP 5000000 dV 174.884 dVAVG 174.884 BOND 217.271 ANGLE
>>> 514.686 DIHED 785.12 IMPRP 0 ELECT -164.908 VDW -207.159 POTENTIAL
>>> 1145.01
>>>
>>> ETITLE: TS BOND ANGLE DIHED IMPRP
>>> ELECT VDW BOUNDARY MISC KINETIC
>>> TOTAL TEMP POTENTIAL TOTAL3 TEMPAVG
>>>
>>> ENERGY: 5000000 217.2714 514.6861 785.1195 0.0000
>>> -3943.6178 -165.2773 0.0000 0.0000 813.3220
>>> -1778.4960 309.5932 -2591.8180 -1768.8754 309.5932
>>>
>>>
>>> I guess the program is not considering the solvation energy in the
>>> calculation.
>>>
>>> Is there any way to correct this?
>>>
>>> Regards
>>>
>>> Felipe
>>>
>>>
>>
>>
>
-- ====================================================================== Thomas Evangelidis PhD student University of Athens Faculty of Pharmacy Department of Pharmaceutical Chemistry Panepistimioupoli-Zografou 157 71 Athens GREECE email: tevang_at_pharm.uoa.gr tevang3_at_gmail.com website: https://sites.google.com/site/thomasevangelidishomepage/
This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:22:48 CST