[AMBER] Using infe = 1, &pmd, and pmemd.cuda (AMBER16)

From: Lenz, Stefan <stefan.lenz.uleth.ca>
Date: Tue, 1 May 2018 10:04:27 -0600

Dear AMBER developers:

I am trying to use the &pmd module with pmemd.cuda (or pmemd.cuda.MPI) on
AMBER16. I have successfully ran my input file using pmemd, however, I get
errors when I try to use the GPU code.

The version of AMBER16 I am using has the patches up to update.11 applied.
According to the manual (Amber17.pdf) what I am trying to do should be
possible: "*From Amber16, we implement these modules from SANDER to PMEMD
and the modules are GPU compatible." *pg. 441.

When I run my input file, I get the error: *invalid reference to variable
in NAMELIST input, unit 5, file user_inputfile.in
<http://user_inputfile.in> line 14, position 7.*

That error points me to my input file:

  ntpr = 2500, ntwx = 5000, ntwr = -125000,
  infe=1 *[line 14]*
  output_file = 'temp_info.dat'
  output_freq = 1
  cv_file = 'cv.in' [*from AMBER manual pg. 447]*

I get the same error using pmemd.cuda, pmemd.cuda.MPI, and pmemd.MPI, but
my input file runs fine using pmemd.

So pmemd.cuda is having trouble parsing infe=1. My question is if &pmd
usage is implemented in pmemd.cuda, and if so whether someone can help me
troubleshoot my input file. If &pmd usage is not implemented in
pmemd.cuda(.MPI) (AMBER16), I was wondering if it is implemented in
pmemd.cuda(.MPI) (AMBER18).

Any help would be greatly appreciated,

AMBER mailing list
Received on Tue May 01 2018 - 09:30:03 PDT
Custom Search