This switches staging buffer allocation for vertex and index data to a linear bump allocation scheme instead of using a pool of buffers. The idea is that for certain apps, vertex uploads tend to be smaller and more spiky. Batching the uploads into a larger shared buffer reduces the number and frequency of buffers we need to request from Metal.
This switches staging buffer allocation for vertex and index data to a linear bump allocation scheme instead of using a pool of buffers. The idea is that for certain apps, vertex uploads tend to be smaller and more spiky. Batching the uploads into a larger shared buffer reduces the number and frequency of buffers we need to request from Metal.