Street Fighter 3: 3rd Strike - 4 Way Real Hardware Speed Comparison!

[hrist]

My biggest issue with Mister is the input configuration.

Real hardware has the IOs directly on the same data bus as the RAM ROMs and other components. mister passes all of that through USB. great that you can plug in a keyboard or game controller but if I'm looking for a real hardware replacement that will never pass muster in my opinion, no matter how great the cores are.

the way to fix this would be to allow the IO directly on the GPIO of the FPGA so the original IO hardware could be implemented in fabric. I'd be surprised if this happened though as it would break the entire accessory ecosystem built around the current serial IO configuration. and only useful if you're playing on a JAMMA cab and have a JAMMA adapter designed to make use of it.

It's been for a while with SNAC and stuff like https://www.ukvac.com/forum/threads/options-for-dual-ram-mister-set-up.87908/
Problem as you say is nowadays this leaves all the i/o logic to the cores and every dev should adapt/code this support.

 

[RynoW]

the way to fix this would be to allow the IO directly on the GPIO of the FPGA so the original IO hardware could be implemented in fabric. I'd be surprised if this happened though as it would break the entire accessory ecosystem built around the current serial IO configuration. and only useful if you're playing on a JAMMA cab and have a JAMMA adapter designed to make use of it.

SNAC behaves essentially as you described, as a lagless interface directly to the FPGA. Jotego cores support them as well, so it should be compatible with the eventual CPS3 core.

The downside is that the behavior is different with every core, so you no longer have the convenience of using the same controller with everything. The creators of BlisSTer tried to square this circle by implementing an API that unified serial behavior across cores, allowing for using any controller with SNAC-like results. But it wasn't merged with main because of its association with a closed source project (the Bliss chips themselves, though other devices could have used LLAPI). Someone is still maintaining LLAPI versions of cores, but BlisSter isn't nearly as popular as it once was now that the ship has sailed on it being merged with main, which is a shame IMO.

At some point though this conversation is purely academic. Is there any practical implication when there are so many USB devices that offer sub-millisecond latency? Once we get to JVS-based systems, those USB devices will have less lag than period-accurate JVS I/O