On Wed, May 30, 2012, Scott Le Grand wrote:
>
> However, pmemd.cuda attains its performance boost by running entirely out
> of cache. With the old approach, 8 * O(m^2) + 4 bytes per atom in the
> cache is consumed as opposed to 8 bytes per atom for storing per atom.
I'm still lost. Don't you need both cn1 and cn2 (or r*, eps)
for each atom (giving 16 bytes per atom), vs. 4 bytes per atom for the
table lookup scheme (plus 8 * O(m^2) for the table itself.) ?
For dhfr, m=17, n=23000, the table only takes 2300 bytes. Factor ix is four
times bigger, but still has only 19 atom types. Cellulose has 400,000 atoms
but only 8 atom types. Generally, the number of types grows only slowly with
system size, and few systems have 50 atom types. So maybe we could check the
number of atom types, and support the full O(m^2) table as long as m is small
enough?
Just a thought on the train...maybe I am missing something here....dac
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed May 30 2012 - 21:00:03 PDT
