Cache calculation of bounding boxes with transforms #3287
Conversation
| std::mem::drop(read); | ||
| let mut write = self.bounding_box_cache.write().unwrap(); | ||
| if write.len() >= 10 { | ||
| write.clear(); |
There was a problem hiding this comment.
Do we really want to clear the cache completely here?
Did you do performance tests on the number of max cache entries, could be interesting.
There was a problem hiding this comment.
I'm using a flush when full replacement strategy which gives the same theoretical guarantees as something like an lru but is significantly easier to implement. If we wanted to improve performance, we could try to remove any scaling factors since they don't affect the outcome and to use fingerprinting to improve the search speed
There was a problem hiding this comment.
Alright, implemented the optimizations
Performance Benchmark Results
|
Improves performance for panning etc.