I have ported the IPS (isotropic periodic sum) code of Xiongwu Wu into
sander
for Amber 9. See amber9/doc/0README.WXW for information. A few notes:
1. I changed the variable IPSOPT to IPS.
2. Self-guided Langevin is not (yet) in Amber 9; volunteers are welcome!
3. Here is the literature reference:
%A X. Wu
%A B.R. Brooks
%T Isotropic periodic sum: A method for the calculation of long-range
interactions
%J J. Chem. Phys.
%V 122
%P 044107
%D 2005
4. There is IPS code in egb(), but it is commented out for now; what is
really needed is IPS for EPOL -- without that, just doing IPS on the
unscreened Coulomb interaction doesn't seem to make much sense (to me,
anyway).
5. Tom Cheatham and I have been running this code (or the equivalent
Amber 8
version) on DNA and RNA, (periodic, explicit solvent simulations). It
seems
to give very good results. Densities are slightly (1%) smaller than with
PME
-- I'm still looking into that.
6. I modified Xiongwu's code some to speed it up, but some more things
are
still needed. There is a square root (in short_ene) that is not done
in a vectorized fashion. Also, the eexips() routine needs to be
parallelized. But it looks like this should indeed scale very well
to
lots of processors. Some current timings are here:
http://amber.scripps.edu/cgi-bin/wiki.pl/Amber_9_Benchmarks
(Look for ips=2 in the top two tables.)
7. Obviously, you should take care in using this code! Try some parallel
runs comparing to regular PME: you should see pretty similar energies
and
structural behavior. You cannot use it
right now for free energy perturbation (icfe=1), but that is a high
priority of mine, since that will be key in further testing the extent
to which this idea can replace PME. Some other combinations of
parameters
may also not work.
8. There are test cases in amber9/test/{gact_ips,gb_rna}
9. Xiongwu: thanks for providing this! Let us know if you have further
experience or ideas. You might think about subscribing to the
amber-developers list: send e-mail to majordomo.scripps.edu, with
"subscribe amber-developers" in the body of the message.
...dac
Received on Wed Apr 05 2006 - 23:49:58 PDT
