Instead of allocating small chunk of GPU memory for separate resources (buffers and textures), sub-allocate memory ranges for resources from a bigger memory heap managed by MTLHeap.