On Thu, Jan 14, 2021, Женя Елизарова wrote:
>
>I have made two practically identical systems (protein + 2 ligands, alanine
>mutant of protein + ligands). During the preparation of the first system, I
>added 40 Na ions and 26 Cl ions, when I tried to add 40 Na ions and 28 Cl
>ions in the second system (charge of the whole system in the first and
>second cases were 0) I had an error: Program received signal SIGSEGV:
>Segmentation fault - invalid memory reference. (sander run) or Error:
>illegal memory access was encountered launching kernel
>kClearForces(pmemd.cuda).
There are tons of reasons why GPU calculations may fail. The usual things
to try: do the minimization and short, intial equilibration on a CPU, then
move to GPU.
It looks like you tried sander: at what stage did the error occur? Please
set ntpr=1 for a short minimization or MD run (like 10 steps) and see if you
see hints about what is going on.
We would need a lot more information to help by doing more than suggesting
debugging strategies.
...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Jan 14 2021 - 14:30:02 PST
