[gmx-users] Re: genconf

[Mark Abraham]

Vitaly V. Chaban wrote:
> Payman,
> 
> Such little charge cannot produce large energies, so it seems you should run
> EM first and everything will be OK with your simulation.
> 
> As far as understand non-zero charge is a resuts of averaging errors (during
> grompp!), you cannot remove this warning.

That's not quite right. It results from the fact that binary floating 
point operations accrue error both from inexact representation of many 
exact decimal values (e.g. 0.02 has no exact representation, and so 
added to itself N times, need not equal N*0.02) and from limited 
precision (to 3 digits, 123 + 0.123 = 123). The former is occurring with 
the sum of atomic charges in grompp. The latter can occur during the 
integration, for example.

Mark