Re: [AMBER] cudaMemcpy GpuBuffer::Download failed an illegal memory access was encountered

From: David A Case <david.case.rutgers.edu>
Date: Fri, 4 Nov 2016 07:58:41 -0400

On Thu, Nov 03, 2016, Andreas Tosstorff wrote:

> ERROR: nfft2 must be in the range of 6 to 512
> ERROR: nfft3 must be in the range of 6 to 512

Amber calculates the size of the reciprocal grid by an automatic procedure,
and the value it comes up with is either too small or too large. What is the
size of your periodic box? Is the size of the y or z dimension greater than
about 500 Ang?

> I attached another complete output file below.

Note sure what you mean by "another" output: did this one fail as above?

> Box X = 89.176 Box Y = 172.506 Box Z = 176.777
> Alpha = 90.000 Beta = 90.000 Gamma = 90.000
> NFFT1 = 96 NFFT2 = 180 NFFT3 = 180

Above looks fine, so I'm wondering if this file gave the errors you report
above? I think we need more information here....

In general, you should equilibrate with constant volume first, before
switching to constant pressure. Try some NVT runs to see if you get similar
problems. It would also be very helpful to know how many MD steps have taken
place when you get your errors.

...dac


_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Nov 04 2016 - 05:00:02 PDT
Custom Search