Generalized abstracted spacial partitioning capability

From BRL-CAD

Space Partitioning - the breaking out of a generic volume of space into structured sub-volumes - is fundamental to a wide range of geometric algorithms. There are multiple methods of partitioning a volume, and while BRL-CAD makes use of partitioning approaches in a number of places with its codebase the partitioning logic is integrated into each individual task and not generalized.

This task would be to design and implement a generalized, abstract method that would allow existing and new code to utilize a pre-implemented partitioning scheme. Say, for example, a ray intersection method wanted to make use of KD-Tree space partitioning - with a successful implementation of a space partitioning API the problem is reduced to implementing calls to test whether nodes satisfy criteria and what action to take in various cases, instead of the details of storing and walking the KDTree data structure itself. Ideally it would be simple for code to swap in and out various space partitioning schemes to compare performance.

Students proposing ideas for this task should discuss it on the email list or IRC to make sure they understand what the requirements for such an API are.

References

  • src/librt
  • src/librt/bool.c
  • src/librt/shoot.c
  • src/librt/prep.c

Requirements

  • Familiarity with C/C++
  • General understanding of space partitioning schemes