CUDA / All Backend Maximum Image Input Sizes

Would anyone be able to direct me to a forum or documentation link that has input image size limits for specifically CUDA backend, but also for the other backends (PVA, VIC, OFA)?

I understand the size limit is algorithm dependent, however the documentation only lists limitations for one or even no backends for many of the algorithms.

For example: VPI - Vision Programming Interface: Convolution tells us that" * On PVA backend, image dimensions must be between 65x33 and 3264x2448." but gives no information about the other backends (what are the CUDA / VIC / OFA input dimension limits for the convolution algorithm?)

As a long-time CUDA programmer, I have no idea what PVA, VIC, or OFA are, and I don’t think they are directly related to CUDA programming and performance. It seems the question might be a better fit for the computer vision and image processing forum:

Ok my apologies, thank you. I will post there.