fujinet #FujiNet - Research Topic - concurrent SIO routine?

[tschak909]

This is a research topic that I hope certain people familiar with SIO routines, such as @HiassofT can chime in:

 

While the current OS routines work great for using N: in many contexts, such as writing turn based games, or communicating with existing services, these routines do not allow for any processing to occur while in use, which occludes their use in applications like arcade games.

 

I do know that there were certain custom loader routines, such as those used by early Lucasfilm games (Ballblazer and Rescue on Fractalus being the two biggest examples), that allowed for two-channel sound use + graphics + logic while disk data was being loaded.

 

I would estimate that the average amount of packet information needing to be sent would be on average about 16 to 32 bytes, to encode changes in game state.

 

Could a new set of routines be developed that could work well in this situation? e.g. have two different routines depending on whether you need to run critical code during these network transfers or not (e.g. DLIs) ?

 

-Thom

Edited September 14, 2020 by tschak909

[xxl]

due to a bug in HV Atari, the IRQ transmission may (and should not) mask the NMI. of course you can modify the operating system to change the order and way of recognizing the IRQ / NMI but this takes memory. I propose to use pooling ... xBios. you're welcome.

 

although ... the procedures for N: on IRQs will work for this application.

Edited September 14, 2020 by xxl