[gmx-developers] stuff in double precision

Berk Hess hessb at mpip-mainz.mpg.de
Mon Feb 5 17:08:42 CET 2007


I am writing this after thinking about David Mobley's bugzilla entry
about differences in energies with and without free energy.
The difference in total potential energy was 0.2 kJ/mol (too big
for David), but the relative error was only 1e-6.

I think this difference only occurs due to summation order differences
of the non-bonded forces, as with free-energy the neighbor lists
are different.

This difference can be avoided by having all energy terms in double
precision. In general this is for free, but it might need recoding
of the assembly loops.
We should consider making most/all variables double precision,
except for the natoms sized arrays.
Also the virial might be more accurate when the shift force
arrays are double precision (I have not checked this though).


More information about the gromacs.org_gmx-developers mailing list