[AMBER] accelerated MD cudaMemcpy GpuBuffer::Download failed an illegal memory access was encountered

From: Ashutosh Shandilya <scfbioiitd.gmail.com>
Date: Fri, 20 Apr 2018 17:59:09 -0400

Dear Amber Users,

I ran MD simulation for 10 ns and then tried to run aMD with following
parameters
---------------------------------------------------------
Accelerated MD - explicit solvent
 &cntrl
  imin=0, irest=0, ntx=1, nstlim=25000000, dt=0.002, iwrap=1,
  ntb=2, cut=10.0, ntp=1, ntc=2, ntf=2, tempi=300.0, temp0=300.0,
  ioutfm=1, ntt=3, gamma_ln=1.0, ig=-1, iamd=1,
 EthreshP=-152417, alphaP=13330, barostat=2, ntpr=5000, ntwr=5000,
ntwx=5000
 &end
-------------------------------------------------------------------------

I encountered with this error while running aMD. My normal MD simulation
run to get parameters also ran with pmemd.cuda which ran fine. Here is the
final output file for that.



      A V E R A G E S O V E R 1000 S T E P S


 NSTEP = 5000000 TIME(PS) = 10100.0 TEMP(K) = 299.95 PRESS
= -1.5
 Etot = -302570.25 EKtot = 73158.56 EPtot =
 -375728.8244
 BOND = 2365.12 ANGLE = 6165.20 DIHED = 9862.4871
 1-4 NB = 2781.82 1-4 EEL = 21490.98 VDWAALS = 43693.4265
 EELEC = -462087.88 EHBOND = 0.00 RESTRAINT = 0.0000
 EKCMT = 32326.40 VIRIAL = 32365.48 VOLUME = 1192465.2387
 Density = 1.0238
 ------------------------------------------------------------------------------

Visual inspection of the pdb structure generated from rst file also looks
normal.


Although, I am not sure why aMD output file shows higher temperature at the
first step

 NSTEP = 0 TIME(PS) = 0.000 TEMP(K) = *437.64*
PRESS = 0.0
 Etot = -58253.34 EKtot = 106742.86 EPtot =
 -164996.20
 BOND = 2353.15 ANGLE = 6119.93 DIHED =
9847.52
 1-4 NB = 2754.62 1-4 EEL = 21776.16 VDWAALS =
 43983.47
 EELEC = -462590.85 EHBOND = 0.00 RESTRAINT = 0.00
 EAMD_BOOST = 210759.76
 EKCMT = 0.0000 VIRIAL = 0.00 VOLUME = 1192520.97
 Density = 1.0238
-----------------------------------------------------------------

Could any one suggest possible solution?

Many thanks,

Ash
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Apr 20 2018 - 15:00:02 PDT
Custom Search