While I love the idea of FPGA co-processors on CPUs, I'm wondering how useful they could be. I guess you could replace the video transcoding unit so you're not tied to one codec but how often do those change anyway.
I must admit I did not think this whole thing fully.
But what was interesting to me was the fact that you could add a peripheral that was not initially intended by the manufacturers, making a mainstream motherboard more versatile.
For you it may be a video transcoding unit, for someone else it may be an SPI or I2C device, PCIe, or extra ethernet, or high quality audio.
I'm not sure what peripherals were implemented by the community for the RP2040 either, maybe they would not make sense on a PC.
Maybe PCIe already does something similar, that's not something I have knowledge about.
Though there is a small difference in my opinion, where, from the point of view of the CPU, it should behave as a normal interface, thus the driver should already exist, and only require a change in the device tree (for linux).
It would still require quite a bit of work:
- The PIO has to behave bug-for-bug compatible with an existing driver
- The exposed pins need have the proper voltage levels & electrical protection
there are a few FPGA dev kits which are set up this way. I have one which has a dual core Atom CPU and a large FPGA connected via PCIe, and the speed is fast.
In that case the CPU is basically the co-processor for the FPGA. I've yet to see a use that wasn't primarily using the FPGA because it needed to be an FPGA, they're not great if you just want to run something fast (outside of a few small uses).
