search

GPUPeople / Ouroboros

GPU MemoryManager based on virtualized queues
MIT License
18 stars 3 forks source link

Biggest chunk queue is the same as a page queue #11

Open delanyinspiron6400 opened 4 years ago

delanyinspiron6400 commented 4 years ago

The biggest chunk queue could use the same implementation as the corresponding page queue, as each chunk only holds one page. The page based implementation is orders or magnitude faster, so that would be a great start.

delanyinspiron6400 commented 4 years ago

Furthermore, the biggest page queue could also reuse its pages aka chunks, as one page is already a chunk. Maybe we could add a custom queue for this, which is a mixture between a page-based and a chunk-based queue.