Right now, we just allocate 1 MiB per frame (soon to be 4 MiB) of CPU-visible GPU memory in order to push constants via dynamic uniform buffer. This number is totally fabricated. We should validate the actual usage, and pick an amount based on that like sane engineers.
Right now, we just allocate 1 MiB per frame (soon to be 4 MiB) of CPU-visible GPU memory in order to push constants via dynamic uniform buffer. This number is totally fabricated. We should validate the actual usage, and pick an amount based on that like sane engineers.