I did't solve the error of sander.MPI, but I found that it works fine with pmemd.cuda. Can I set the nmropt=1 and use the pmemd.cuda to do it?
By the way, what's the difference between sander and pemem?
> -----原始邮件-----
> 发件人: "David A Case" <david.case.rutgers.edu>
> 发送时间: 2021-04-21 22:28:16 (星期三)
> 收件人: "AMBER Mailing List" <amber.ambermd.org>
> 抄送:
> 主题: Re: [AMBER] running sander.MPI :forrtl: severe (174): SIGSEGV, segmentation fault occurred
>
> On Wed, Apr 21, 2021, 周丽萍 wrote:
>
> >I have set ntpr=1,but the out file has nothing. The attachments are mdout
> >files, anneal6.out for sander.MPI and anneal5.out for sander.
> >
>
> OK: you need to continue debugging:
>
> (I'm assuming that you have run the parallel tests for sander. If not, be
> sure to do that, unless you know that other jobs using sander.MPI run fine.)
>
> * Try just two processors, not 32.
> * You are not using anything (that I see) that needs nmropt=2, so set that
> to 1
> * Turn off the nmr stuff completely, to see if you can get a simple
> calculation to run. (Just set nstlim=10 or something small: you are
> trying to find the simplest system that fails.
>
> ....dac
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
>
http://lists.ambermd.org/mailman/listinfo/amber
</amber.ambermd.org></david.case.rutgers.edu>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Apr 21 2021 - 22:30:02 PDT
