High complexity and relatively low reward. There have been many attempts at these sorts of dual architectures through history, and they've mostly been abject failures for various reasons (there have been some dual PC/Mac designs from the 90s, those Sun PCI cards with 80386s built on them, etc)
At the low end of complexity, you could put a raspberry pi on a pcie daughterboard and just power it off the slot, and maybe fake a NIC to the host for file transfer.
That's not really any more useful than just having a separate, dedicated raspberry pi though. To be useful, you need to implement some sort of deeper synchronization. Perhaps you could have a shared memory aperture between the two systems, a virtio style interface (but physical, not virtual!) with the associated protocols for passing through files, visuals, etc.
But it's all complicated and expensive, and in the end unless you're doing CPU level debugging and need to sync a debugger with low level CPU state, there's really not much advantage besides some space saving to just having a separate x/y/z on your desk.
The reverse was also true; I worked for a company back in the 1990's that provided third part SPARC solutions - probably the only company in the Southern Hemisphere at the time that did so.
One of the more interesting products we resold were the ROSS Technology "Sparcplugs" - you basically had a SPARC motherboard and CPU that fit into a couple of 5.25" drive bays on a host PC. So almost the reverse of the SunPCI cards.
At the low end of complexity, you could put a raspberry pi on a pcie daughterboard and just power it off the slot, and maybe fake a NIC to the host for file transfer.
That's not really any more useful than just having a separate, dedicated raspberry pi though. To be useful, you need to implement some sort of deeper synchronization. Perhaps you could have a shared memory aperture between the two systems, a virtio style interface (but physical, not virtual!) with the associated protocols for passing through files, visuals, etc.
But it's all complicated and expensive, and in the end unless you're doing CPU level debugging and need to sync a debugger with low level CPU state, there's really not much advantage besides some space saving to just having a separate x/y/z on your desk.