GitLab

Fixed an edge case bug for Gyrcuk Radii. If the radius (ri) of the atom being descreened is larger than sum of the separation distance (r) and scaled descreening radius (sk), then there is no descreening. This could happen for carbon atom (ri ~= 2) engulfing its bonded hydrogens (sk < 1).

o Variable renaming in multipole electrostatics code, to avoid confusion.
o CUDA test case only runs finite difference test for the double precision test.

errors.

This copies the fix for the AmoebaTorsionTorsionForce over to the other two
forces that calculate angles via the inverse cosine -- specifically the in-plane
angle force and the stretch-bend force.  The TorsionTorsion fix did not
completely solve the occasional simulation blowup, although it did make it a
*lot* less frequent (it went from happening within 5 minutes to happening ~once
per day).

of just 1.  Backwards compatibility is provided by making the 2nd force constant
default to -1, which is reinterpreted as "copy the first force constant".

Updates both the reference and CUDA kernels.