Prefetch one more word

[larsbrinkhoff]

More completely model the IRC/IR/IRD prefetch queue.

[stefanberndtsson]

Can this be done in a way compatible with the old cpu code?

[larsbrinkhoff]

Yes, I think so. I started doing that a week ago, but ran into a dead end.

But there's a great advantage in having the old instructions work more correctly, because they're mixed with the new instructions.

[larsbrinkhoff]

Basically, all instructions that do a bus_read_<byte/word/long> from cpu->pc, have to be changed to use fetch_instr instead.

Also, there's some trickery going on with cpu->pc being incremented or not. I'm guessing that the old ea_read functions can be part of a read-modify-write operation, in which case the PC should be left untouched until the ea_write is done?

[larsbrinkhoff]

One possible shortcut would be to have all old instructions clear out the prefetch queue at the beginning, and fill it at the end according to the value of cpu->pc.

This quick-and-dirty hack would only be to provide an execution environment for the new clocked instructions.

[larsbrinkhoff]

By the way, when I wrote a VAXstation 100 emulator, I wrote a 68000 engine that models the prefetch queue. https://github.com/larsbrinkhoff/absolutely-not-a-vaxstation100-emulator/