[gmx-developers] parallelization

[David van der Spoel]

Hi,

I didn't copy alll Justin's comments. Would be too much....
I like the documentation page, it definitely shows some logic in the 
datatypes at least, most of those not starting with t_ are actually 
imported source code, like DSSP. Also the cross-referencing is nice (which 
routine uses what).

As to your MC plans, it sounds ambitious! I actually have just merged 
nmrun and xmdrun into mdrun to get rid of some of the extra overhead. 
Duplicate code provided quite a bit of maintainance nightmares in the past 
and I just fixed a bug in congrad due to not duplicating code correctly.
(So Erik, let's keep it lagom). I would prefer to have the integrators in 
update.c (if they are of general interest) and just make it easier to have 
a driver program call these functions. Simpler interfaces. We will also 
need some kind of (simplified) call-tree for mdrun. Can doxygen do that?

Anyway, before you start working on something like that, try to separate 
the physics from the implementation. Think of what you want to do before 
you start implementing.


-- 
Groeten, David.
________________________________________________________________________
Dr. David van der Spoel, 	Biomedical center, Dept. of Biochemistry
Husargatan 3, Box 576,  	75123 Uppsala, Sweden
phone:	46 18 471 4205		fax: 46 18 511 755
spoel at xray.bmc.uu.se	spoel at gromacs.org   http://zorn.bmc.uu.se/~spoel
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++