[gmx-users] gromacs 4 release date?
David van der Spoel
spoel at xray.bmc.uu.se
Sat Jan 6 10:59:07 CET 2007
Chris Neale wrote:
> Any projection date for a release? Of particular interest to me is
> whether it is very likely to come out before Jan 1 2008.
I hope we can release it before summer, but it depends on what we decide
to put in it besides the new parallelization. Since it will be a major
upgrade, we don't want to force people into upgrading again, too soon
after the first release.
> I have heard that v4 is expected to scale better for parallel runs than
> previous versions, is this correct? If so, how close to the namd scaling
> might gromacs get?
Scaling is a bit of a problem for comparison, as you can make your code
scale better by making the sequential bit slower. Berk has made the
standard GROMACS benchmark for DPPC to scale to well (> 80%) over 100
CPUs on an AMD opteron cluster with infiniband, that means below 1000
atoms/CPU still scales very well. I think we can get the same absolute
performance as NAMD at 20-40% of the number of processors.
Even there, absolute performance is hard to compare since it is not
straightforward to really get all the settings identical. E.g. NAMD
(AFAIK) typically uses double precision versus single in GROMACS, but
NAMD does PME only every fourth step while GROMACS does it every step.
Note that you can compile GROMACS in double precision, which will make
it slower and hence scale better...
David van der Spoel, PhD, Assoc. Prof., Molecular Biophysics group,
Dept. of Cell and Molecular Biology, Uppsala University.
Husargatan 3, Box 596, 75124 Uppsala, Sweden
phone: 46 18 471 4205 fax: 46 18 511 755
spoel at xray.bmc.uu.se spoel at gromacs.org http://folding.bmc.uu.se
More information about the gromacs.org_gmx-users