If we have many room templates, a non-trivial amount of time can be spent to compute the configuration spaces. For example, in the Enter the Gungeon example, it takes more than 1 second to do so. One possible solution to this problem is to somehow cache the results (if we are not able to significantly improve the performance of the computation).
Random ideas:
the caching layer should be mostly transparent for users
we have to make sure that the cache is not too large (maybe use binary serialization?)
we have to somehow identify individual room templates (polygon + door positions?)
to save some space, we can first add ids to individual room templates and then identify configuration spaces pairs with these ids
Next steps:
[ ] Check how large would be a very simple representation of the cache
A partial solution would be to at least cache intermediate results during runtime to make sure that nothing is computed more than once.
If we have many room templates, a non-trivial amount of time can be spent to compute the configuration spaces. For example, in the Enter the Gungeon example, it takes more than 1 second to do so. One possible solution to this problem is to somehow cache the results (if we are not able to significantly improve the performance of the computation).
Random ideas:
Next steps:
A partial solution would be to at least cache intermediate results during runtime to make sure that nothing is computed more than once.