Hi,
In addition to Dave's advice, if you're going to use the CUDA code for
minimization you should use the full double-precision code since it
will be less likely to overflow with very large initial forces. If you
still have issues, your best bet is to try sander, which although
slower tends to be more robust for minimization and will often (but
not always) provide better error messages.
-Dan
On Sat, Oct 12, 2024 at 10:21 AM David A Case via AMBER
<amber.ambermd.org> wrote:
>
> On Thu, Oct 10, 2024, Ryan Woltz via AMBER wrote:
> >
> >http://archive.ambermd.org/202106/0032.html
> >
> >I am getting this error during minimization:
> >Program received signal SIGSEGV: Segmentation fault - invalid memory
> >reference.
>
> I don't have a solution per se, but try running minimization (with many
> fewer steps, say 500 to 1000) with sander using ntmin=3.
>
> The minimization code in pmemd is not very robust, and sander will do a much
> better job. It will indeed be slower, but most systems don't require a lot
> of minimization steps prior to MD.
>
> >While I'm here I also do have an issue with a single channel simulation
> >that seems to be relatable...
>
> I don't have any ideas here. The error is indeed related to CUDA.
>
> ...dac
>
>
> _______________________________________________
> 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 Thu Oct 17 2024 - 06:30:02 PDT