jwlawson / ptope Goto Github PK

Best chances of finding a large polytope is by explicitly trying to close off the faces which were not finished.

Haven't really thought through the process yet.

Current tests are very limited. Should be expanded.

Currently each call causes an allocation of a vector<pair>, which should be avoided.

Something like bool has_compatible( index ) which checks if there are any vectors compatible with with the specified one.

For best performance it's possibly best not to create a dynamically linked library, but to make a static one which would enable me to use link time optimisations in the MPI program.

I'm not too sure whether I can do both (which library would the linker choose?) or if I even need to worry about that. I'm also not too sure how much of an impact the LTO would give inlining functions.

Need some way of storing which vectors are pairwise compatible.

The original idea was an array of arrays, where a vector would map to a list of the vectors compatible with it. This uses a lot of memory, and perhaps it would be better to use something like a bitset to provide a smaller footprint at the cost of slightly more computations (perhaps).

Current test only uses next_compatible_to(0, next), which seems to hide a bug for higher indices.