Does this "dynamic caching" technique go by another name, as a generic concept? I'm having trouble finding information about it and why they are saying it hasn't been done before.
sound very plausible, potentially spot on.
Also seems like this would have to be something "opt in" or else what stops someone from just using GL/Vulkan/Metal to allocate and lock a giant buffer every frame anyway?
I'm optimistic, but I am really eager to see the documentation.
If all that is true, it sounds like a good improvement. And it sounds plausible that Apple was the first since it would require the whole system to work in synch without introducing latency in rendering, a good use of an SoC. I’d still like to see benchmarks and understand potential failure/bottleneck scenarios.