Concurrent Chunk Management Engine

Concurrent Chunk Management Engine

2M Downloads

Suggestion for faster handling of across chunk structures/features

Mr-Squid42 opened this issue ยท 1 comments

commented

since c2me is designed for high concurrency, something like this is likely already implemented, but I'm going to suggest it anyway

Suggestion:

  • use Scatter-as-Gather for handling cases where, for example trees need to generated across boundaries, this is done by:
  • instead of using scatter operations, like using one thread to write out to multiple location requiring mutiple memory and locks, data access is instead decentralized, where each chunk gathers block data from surrounding chunks
  • they then work out possible places where trees can generate across boarders independently and generates the trees, if any blocks land in the chunk gathering they are then written in
  • this video may better explain it better than i can (13:45): https://www.youtube.com/watch?v=2mpklUE7BfA&t=857s

Reasons:

  • this level of parallelism could speed up terrain generation a lot if not already implemented
commented

Not going to be implemented any time soon. The current model works fine.
Using that approach would require large refractors and is going to break compatibility.