Hi AMBER community,
I'm currently running temperature-based replica exchange on AMBER for a
soluble protein system using an explicit solvent.
In certain instances, after the replica exchange simulation has been
running smoothly for a good number of steps and exchanges, I suddenly hit a
GPU error:
*Error: an illegal memory access was encountered launching kernel
kClearForces*
*cudaFree GpuBuffer::Deallocate failed an illegal memory access was
encountered*
It seems that this error has been observed in the past, but is the
source/cause known?
For your reference, I use the following input parameters for my t-REMD
simulation:
&cntrl
imin=0, ! Molecular dynamics
ntx=5, ! Positions and velocities read formatted
irest=1, ! Restart calculation
ntc=2, ! SHAKE on for bonds with hydrogen
ntf=2, ! No force evaluation for bonds with hydrogen
WATNAM='TIP3',
OWTNM='OH2',
tol=0.0000001, ! SHAKE tolerance
ntt=3, ! Langevin dynamics
gamma_ln=1.0, ! Collision frequency for Langevin dyn.
temp0=XXXXX, ! Simulation temperature (K)
ntpr=50000, ! Print to mdout every ntpr steps
ntwr=50000, ! Write a restart file every ntwr steps
ntwx=50000, ! Write to trajectory file every ntwc steps
dt=0.004, ! Timestep (ps)
ig=-1, ! Random seed for Langevin dynamics
cut=9.0, ! Nonbonded cutoff (Angstroms)
ioutfm=1, ! Write binary NetCDF trajectory
ntxo=2, ! Write binary restart file
barostat=2, ! Use Monte Carlo barostat (Amber 14)
ntp=0, ! No pressure coupling, not supported by AMBER REMD
ntb=1, ! Constant volume simulation, NVT
nstlim=5000, ! Number of MD steps between each exchange attempt
numexchg=1000, ! Number of exchange attempts during simulation
nmropt=1, ! Read in chirality restraints
/
&wt TYPE='END'
/
DISANG=$RESTRAINTS_FILE
Many thanks once again,
Jason
--
Jason Wang
Stanford University, Class of 2018
jwang198.stanford.edu
ᐧ
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Apr 05 2018 - 14:00:03 PDT
