AMBER Archive (2008)

Subject: Re: AMBER: Compilation of amber8 patched with ramd

From: Vlad Cojocaru (Vlad.Cojocaru_at_eml-r.villa-bosch.de)
Date: Wed Jun 25 2008 - 04:00:10 CDT

Hi Francesco,

Happy to hear that you managed to compile the RAMD. Your test looks fine
to me for now. I would use "iramdbug=2" in your inputs to get as much
output as possible. Also, please do a serial test as well and compare
the force values between runs on different number of processors. As far
as I remember we had some troubles with running the RAMD serial. Since I
only run in parallel, I dont know if that was fixed or not. I recommend
this test just to be sure before starting the production runs. In case
you do get significant differences between the serial and parallel runs,
make sure that parallel runs on different number of CPUs produce same
results and let me know.

Please take the input values from the test as "not realistic"... do
systematic tests to chose your values. The important point is to have
reasonably long trajectories (at least 1-5 ns per run in function of
your resources) with the smallest possible perturbation. Make sure your
system does not distort. If system distorts it is a sign that decreasing
the "accel" value is required. Decreasing accel (or the number of RAMD
steps) has more or less similar effect on the trajectory length as
increasing the "thresh" value (but the cause is different).

Please also try to see if the combined RAMD-MD brings any benefits (from
my tests it is not clear yet but it should dump the perturbation on the
system and it should increase the sampling along the path). I have some
plans to write a communication about RAMD for the purpose of general
usage and reproducibility but before that I need to get my current
results published and then I can concentrate on doing all further tests
required for such a communication.

Also, I recommend Langevin dynamics for temperature control. Although we
have used Berendsen till now, latest tests I have done indicate that LD
is probably a better choice (especially since RAMD already introduces a
random force on a small part of the system). But of course you can test
the thermostats yourself if you wish.

Good Luck
vlad

Francesco Pietra wrote:
> This is especially directed to Vlad and Stefen, as they know ramd:
>
> For the parallel compilation I made recourse to tricks as the flag
> -openmpi was not recognized. Having MPICH_HOME pointing where MPI_HOME
> is (although I have only openmpi) I used the flag -mpich.
>
> The test "make test.sander" was met with failure for tgtmd only. I
> believe it can be forgotten.
>
> The test ./Run.ramd (while still having the env var DO_PARALLEL for 4
> processors) only reported some diff that seem to me acceptable given
> the different platform (the serial test was completed in 5 sec against
> the 14 sec used in 2005 on your platform). Could you please have
> acursory view at the attached files to confirm my impression and
> giving me the imprimatur to use ramd? Thanks (thousand)
>
> francesco
>
>
>
> On Wed, Jun 25, 2008 at 12:06 AM, David A. Case <case_at_scripps.edu> wrote:
>
>> On Tue, Jun 24, 2008, Francesco Pietra wrote:
>>
>>
>>> Running the general serial test for amber8, all PASSED except
>>> newton_raph (forrtl: severe (174) segmentation fault). That likely
>>> because no bugfix was applied (and it can't) to amber8 before patching
>>> with ramd. I was unable to trace what newton_raph is expected to do,
>>> so that may I ask whether this is a problem for running ramd?
>>>
>> As I understand it, RAMD is a modification to *sander*, and newton_raph
>> is an *nmode* test. They are complete separate executables.
>>
>> ...dac
>>
>> -----------------------------------------------------------------------
>> The AMBER Mail Reflector
>> To post, send mail to amber_at_scripps.edu
>> To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
>> to majordomo_at_scripps.edu
>>
>>
>
>
>
> 

-- 
----------------------------------------------------------------------------
Dr. Vlad Cojocaru

EML Research gGmbH
Schloss-Wolfsbrunnenweg 33
69118 Heidelberg

Tel: ++49-6221-533266
Fax: ++49-6221-533298

e-mail:Vlad.Cojocaru[at]eml-r.villa-bosch.de

http://projects.villa-bosch.de/mcm/people/cojocaru/

----------------------------------------------------------------------------
EML Research gGmbH
Amtgericht Mannheim / HRB 337446
Managing Partner: Dr. h.c. Klaus Tschira
Scientific and Managing Director: Prof. Dr.-Ing. Andreas Reuter
http://www.eml-r.org
----------------------------------------------------------------------------

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
      to majordomo_at_scripps.edu