[gmx-developers] single precision, double precision hybrid GROMACS

[D. Ensign]

Greetings,

I have been charged with allowing some quantities (most specifically, the forces) to be
computed in single precision, while the rest are computed in double precision. We hope
this will allow us to get the best precision for the least computational cost. There are
several ways I can think of implementing this, and I list them here for your insightful
criticism and heartless ridicule (you can ridicule nicely if you want, too):

1. starting from the single precision point of view, change certain important calculations
to double precision -- say, start with the constraints (then maybe making other quantities
double, later on); or,

2. start with double precision and make all the force calculations occur in single
precision; or,

3. compiling two libraries, one in double precision and one in single; use, eg, the single
precision version of fnbf.o and double precision integrator, etc.

What are your thoughts?

Thanks,
Dan Ensign
Stanford University
-- 
"Give up learning, and put an end to your troubles." -- Lao Tzu