Hi Marc,
Yes, Amber 12 GPU does NOT support off diagonal modification to the VDW
terms. Prior to bugfix.9 this was not tested for so the code will run but
the answers are incorrect. From bugfix.9 onwards the code will quit with
the error message you see below.
Amber 14 GPU supports the off-diagonal modifications so you will need to
update to this. Backporting it to AMBER 12 would be substantial work.
All the best
Ross
On 8/8/14, 4:12 AM, "Marc van der Kamp" <marcvanderkamp.gmail.com> wrote:
>Dear amber-ites (and particularly the pmemd.cuda developers),
>
>I have generated a prmtop with chamber. Subsequently, I modified the
>prmtop
>with parmed.py to change the LJ parameters in accordance to the NBFIX
>corrections (for sodium - carboxylate interactions) that are now
>recommended for the CHARMM protein force-field (v36).
>
>sander & pmemd have no problem running with this modified prmtop, but
>pmemd.cuda gives the following error:
>
>CUDA (GPU): Prmtop appears to contain modified off-diagonal elements for
>VDW A coef.
> GPU code does NOT support such force field modifications.
> Found Aij = 61596.8299420842 expected
> -96740.2772678219
> i = 36 j = 3 Matrix element = 633
>
>(I've confirmed that pmemd.cuda runs fine with the prmtop prior to the
>NBFIX correction.)
>This is NOT the most recent pmemd.cuda version (I don't have access to
>that
>yet), I used
>Version 12.3.1
>
>Can I confirm that it is thus not possible not to use the NBFIX
>corrections
>in combination with pmemd.cuda?
>And, are there any plans to support this in the future? (Or is it already
>supported in a more recent pmemd.cuda version?)
>
>Many thanks,
>Marc
>_______________________________________________
>AMBER mailing list
>AMBER.ambermd.org
>http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Aug 08 2014 - 09:00:02 PDT
