On Apr 26, 2008, at 1:28 PM, Ben Webb wrote:
> Daniel Russel wrote: >> On Apr 26, 2008, at 10:56 AM, Ben Webb wrote: >>> I don't think we can use CGAL. It is certainly available for Fedora, >>> but >>> we build on a lot of other platforms too. >> It is tested on quite a variety of platforms. More than IMP, but I >> have not checked if it is a superset or not. > > Indeed, but I think you missed my point: in order for us to test IMP > with CGAL on all the platforms we build on, somebody (i.e. me) has to > install CGAL on those platforms. I'm reluctant to do that unless we > really need it. What do others think? Does anybody need CGAL for > anything else? Requiring it just for a nonbonded list seems overkill. >
CGAL might be very useful for various tasks such as grid manipulation, motion planning, dimensionality reduction and more. However, I really do not think it should be part of IMP kernel , which should provide basic functionality common to most tasks.
Two points: 1. We should have definition of what goes or does not go into the kernel - I think that now pretty much everything is part of the kernel ( other the embed lib) 2. Daniel - can you please elaborate more on the algorithmic differences between your "improved implementation" and the CGAL one? I can not help wondering why a simple geometric hashing (~5 c++ classes) can not help here.