Syphon-Framework
Syphon-Framework copied to clipboard
Rethink direct CPU memory access limitation due to unified memory model on new systems?
Given that M1 machines provide universal memory and GPU shares memory with CPU, it might make sense to revisit our 'hiding' of the raw IOSurface in Syphon?
Just branching the subject to discuss w @bangnoise
We could in theory provide an IOSurface backed CVPixelBufferRef as a method in the framework? Thinking aloud? I imagine this would open up some use cases and provide a unified way to access surfaces with a clean interface.
I've got a good use case of 'getting' a new IOSurface from some other library. It would be cool to be able to publish using either the SyphonServerBase
class, or a new SyphonSurfaceServer
which consumed an input IOSurface and polished it directly?