Hello,
I tried to run the same simulation using double precision on the GPU and
the output seems to be fine, whereas the single precision seems to throw a
bunch of errors (coordinate overflow etc.). Of course, the simulations with
double precision pmemd implementation runs a lot slower than what we would
expect a GB simulation to run. Is there a way/combination of parameters in
the MD input file that could be run for the single precision implementation
of pmemd?
Here is the input file used:
&cntrl
ntwx = 100,
ntwe = 100, ntwr = 50, ntpr = 1000, ntxo = 1, ioutfm = 0,
imin = 0,
ntx = 1,
temp0 = 300,
tempi = 300,
ig = -1,
nstlim = 20000,
dt = 0.002,
ntb = 0,
nscm = 100,
ntc = 2, ntf = 2,
ntt = 2, vrand = 1000,
igb = 1, cut = 999.0,
nmropt = 1,
ntr = 1,
restraintmask = ':1-X_LASTREC_RESIDNO_X & !.H= ',restraint_wt = 100,
/
&wt type='END', /
DISANG=./restraints.in
My restraints.in file looks like:
&rst
iat=1620,1625,
ialtd=1, r1= 0, r2= 0.1, r3= 15, r4= 15.1,
rk2=1.5, rk3=1.5,
/
Additionally, if I set igb = 1 and cut < 999.0, pmemd does not run. The
manual says "igb /= 0 && cut < systemsize '' is not supported in GPU. What
is the system size being referred to here?
I would appreciate it if someone can shed light on this.
Thank you.
On Sat, Nov 5, 2022 at 2:01 PM Prithviraj Nandigrami <
prithviraj.nandigrami.gmail.com> wrote:
> Thank you for the suggestion. What I seem to be confused about is that
> some of the ligands (Arginine for example) run without flaw every
> combination I have tried, whereas for other ligands (CYS, HIS for example)
> seem to be throwing these errors frequently.
>
>
> On Fri, Nov 4, 2022 at 7:30 PM Thomas Cheatham <tec3.utah.edu> wrote:
>
>>
>> > However, it still shows the message about coordinate overflow as
>> before. The five minimization steps run fine. The
>> > error occurs in the 06md.in script (stage).
>> >
>> > Let me know what you think I should try next.
>>
>> As mentioned, print every step mdout and trajectory; look at mdout,
>> visualize trajectory... --tec3
>>
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Nov 08 2022 - 15:30:03 PST
