[AMBER] CUDA and Restraint Error

From: Kung, Ryan <ryan.kung.uleth.ca>
Date: Fri, 4 Jan 2019 14:51:18 -0700


I am having trouble running restrained MD simulation with pmemd.cuda on
AMBER18. I saw a similar error in the archive but it doesn't seemed to be
fixed (http://archive.ambermd.org/201604/0527.html#options2). Specifically,
when I attempt to run pmemd.cuda on GPUs it fails in the first 10000 steps
but it works fine when I use the pmemd.MPI code on CPUs. I have attempted
on two different types of GPUs but I get the same error.

The error istelf invloves the energy in the out file going to ******* and
the restraint distance jumping unrealistically. One example is the COM
distance for my system went from 28A to 328A in one step. My system is not
even 100A across so I don't know how this is happening.

My in file looks like:

production: 45 ns simulation
  nstlim=22500000, dt=0.002, ntx=5, irest=1,
  ntpr=1000, ntwr=2500000, ntwx=1000,
  ntt=3, temp0=310.0, gamma_ln=3.0, ig=-1,
  ntp=1, pres0 = 1.0, taup=2.0,
  ntc=2, ntf=2, iwrap=1, ntb=2,
  cut = 10.0, ioutfm = 1,
 &wt type='DUMPFREQ', istep1 = 50,
 &wt type='END',

and my restraint file is:

 &rst iat=-1,-1, r1=18.6, r2=28.6, r3=28.6, r4=38.6, rk2=10., rk3=10.,
ir6=0, ialtd=0, iresid=0, igr1=6229,6251,6252,6231,6294,6296,
igr2=9515,9516,9519, /

Does anyone have any insight to what is wrong. Thanks.

Ryan Kung
AMBER mailing list
Received on Fri Jan 04 2019 - 14:00:01 PST
Custom Search