Re: At high cutoff: "ERROR: CALCULATED VDWA FORCE MAY NOT MATCH ENERGY! POSSIBLE BUG!"

From: Brian Radak (brian.radak_at_gmail.com)
Date: Sun Jul 01 2018 - 12:21:03 CDT

This indeed looks like a problem with precision in the interpolation grid
of the Lennard-Jones code. Unfortunately the VdW and electrostatic cutoffs
are always the same to my knowledge. That being said, it is extremely
unlikely that a large cutoff is more tractable than PME or MSM. The former
must scale as N^2 while the latter are roughly linear scaling.

How big is your system? Users regularly use PME on order one million atoms.

On Sat, Jun 30, 2018, 4:37 PM Victor Kwan <vkwan8_at_uwo.ca> wrote:

> Hello all,
>
> I am trying to investigate whether long range electrostatic interactions
> changes the result of my calculation. Unfortunately MSM and PME is not
> an option for my system. When setting cutoff to 400 Angstrom, I've found
> the following in the log file:
>
> Info: ABSOLUTE IMPRECISION IN FAST TABLE ENERGY: 1.65436e-23 AT 399.68
> Info: RELATIVE IMPRECISION IN FAST TABLE ENERGY: 5.30522e-17 AT 399.68
> Info: INCONSISTENCY IN FAST TABLE ENERGY VS FORCE: 0.000290475 AT 0.251946
> Info: ABSOLUTE IMPRECISION IN VDWA TABLE ENERGY: 2.53164e-47 AT 399.68
> Info: RELATIVE IMPRECISION IN VDWA TABLE ENERGY: 2.63784e-15 AT 397.109
> Info: ABSOLUTE IMPRECISION IN VDWA TABLE FORCE: 5.34553e-51 AT 397.109
> Info: RELATIVE IMPRECISION IN VDWA TABLE FORCE: 1.74517e-16 AT 397.109
> Info: INCONSISTENCY IN VDWA TABLE ENERGY VS FORCE: 1.15649 AT 398.397
> ERROR:
> ERROR: CALCULATED VDWA FORCE MAY NOT MATCH ENERGY! POSSIBLE BUG!
> ERROR:
> Info: ABSOLUTE IMPRECISION IN VDWB TABLE ENERGY: 9.86076e-32 AT 399.68
> Info: RELATIVE IMPRECISION IN VDWB TABLE ENERGY: 6.31555e-17 AT 399.68
> Info: ABSOLUTE IMPRECISION IN VDWB TABLE FORCE: 2.40741e-35 AT 397.109
> Info: RELATIVE IMPRECISION IN VDWB TABLE FORCE: 1.933e-16 AT 397.109
> Info: INCONSISTENCY IN VDWB TABLE ENERGY VS FORCE: 2.55495 AT 398.397
> ERROR:
> ERROR: CALCULATED VDWB FORCE MAY NOT MATCH ENERGY! POSSIBLE BUG!
> ERROR:
>
> I understand that NAMD uses interpolation table to calculate force and
> energy. But what is the implication here? Is the energy output (KINETIC,
> TOTAL, TOTAL2 etc.) _potentially_ unreliable? I assume the imprecision
> is caused by floating-point error.
>
>
> Best wishes,
>
> Victor Kwan
>
>
>
>

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2018 - 23:21:13 CST