Dear AMBER users,
I am trying to pull a group of molecules (ca. 2000 atoms) to a lipid
bilayer (ca. 12000 atoms)
by the steered MD using AMBER16. The distance between the group and
bilayer is defined by
the center-of-mass distance of them.
According to the manual, the number limit of atoms that can be used to
group a center of mass
is 1024 (pmemd). So, I changed "maxigr = 1024" to "maxigr = 13000" in
nmr_calls.F90.
This worked well for pmemd.MPI.
However, when I used pmemd.cuda.MPI, the calculation halted with the
following errors.
Error message:
cudaMemcpy GpuBuffer::Download failed an illegal memory access was
encountered
MDOUT:
NSTEP = 0 TIME(PS) = 0.000 TEMP(K) = 0.00 PRESS
= 0.0
Etot = ************** EKtot = 0.0000 EPtot =
**************
BOND = 6713.2488 ANGLE = 22705.2587 DIHED =
13265.5840
1-4 NB = 6053.1327 1-4 EEL = 75505.7247 VDWAALS =
15747.5350
EELEC = -366586.2375 EHBOND = 0.0000 RESTRAINT =
**************
EAMBER (non-restraint) = -226595.7535
EKCMT = 0.0000 VIRIAL = 0.0000 VOLUME =
1138225.3717
Density =
1.0378
I suspected a collision due to the restraint. I tried rk2 from 7.2 to
0.0 and conducted
the simulation again. However, the result was the same.
How can I use the maxigr greater than 1024 with pmemd.cuda.MPI?
The MDIN and DISANG files are as follows:
MDIN:
ASMD simulation
&cntrl
imin = 0, nstlim = 500000, dt = 0.002,
ntx = 1, temp0 = 303.0,
ntt = 3, gamma_ln=1.0 ntc = 2, ntf = 2,
ntwx = 1000, ntwr = 500000, ntpr = 1000,
cut = 10.0, ig=-1, ioutfm=1,
irest = 0, jar=1,
ntxo=2, barostat=2, ntb=2, ntp=2,
/
&wt type='DUMPFREQ', istep1=1000 /
&wt type='END' /
DISANG=dist.RST.dat.1
DUMPAVE=asmd_1.work.dat.1
LISTIN=POUT
LISTOUT=POUT
DISANG:
&rst
iat=-1,-1, fxyz=0,0,1,iresid=0,
igr1=1609,1610, *snip*, 17005,17006,
igr2=1,2, *snip*, 17151,17152,
r2=42.123, r2a=38.123, rk2=7.2,
&end
Best regards,
Shuntaro Chiba
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jun 08 2016 - 08:30:02 PDT
