Suballocate Buffers on Metal #3109
Labels
area: performance
How fast things go
backend: metal
Issues with Metal
help required
We need community help to make this happen.
type: enhancement
New feature or request
Milestone
Similar to #2720, we should consider using gpu-alloc to sub-allocate
MTLHeap
s instead of just creating raw buffers every time. Unlike DX12 though this isn't as obvious of a perf win because metal has a "automatic memory management" lens to it to some APIs, so the change should be benchmarked. It would also be good to consult metal literature to see if MTLHeap is a performance thing or not.The text was updated successfully, but these errors were encountered: