I have been watching with great interest the SR-IOV videos. And I was very happy to see the latest video on the issue that dropped. I have a lot of questions though. Since my field is cyber security, the questions are tilted a bit to the side.
-
How does SR-IOV work, from a memory perspective? Is there any good documentation in the field. More specially how well are the VF:s isolated from each other. Can PF read VF:s memory.
-
How can you configure the VF:s An NVIDIA card of modern vintage has over 10 000 CUDA cores, can you assign say 500 CUDA cores to one VF and then 1000 to the next. Is it the same thing with memory, can I give 1GB to one VF and 2GB to the next? How does this look for other vendors, AMD has 80 cores in there top of the line GPU, Intel ARC has 32? cores in its setup? As far as I can tell an Intel 13900k has a iGPU with 32 cores. How about ports, can you assign different displays to different VF:s?