[gmx-users] question about Polak Ribiere conjugate gradient

[Erik Lindahl]

Low-memory Broyden-Fletcher-Goldfarb-Shanno (guess why we use the 
acronym :-)

This is a variable metric method I implemented for 3.2; it is usually 
more efficient than CG, but it involves an iterative step that we 
haven't bothered to parallelize. CG is significantly improved in 3.2, 
too (and it works in single too, now).


> BTW, what integrator is that? Is such new integrator present in GMX
> Manual?  Is there more, like Newton-Raphson?

I'm pretty sure I added it to the manual, but tell us if it's not 
there. Both CG and BFGS are Newton-style generalized methods for 
multiple dimensions.

Cheers,

Erik