The server seems to have vanished from the face of the earth.
I’ve finally purchased my own domain and hosting so a new (stable) site should be online in a week or two. I’ll keep you posted.
Regarding the 68k emulator, I’ve made fair progress with it and finally went for a code generator: It generates the code for all opcodes at compile time. I’m guessing this would give best performance but it makes development a bit slower because:
a) if I do a bugfix I have to re-generate the sources again.
b) because the huge number of opcodes, the sources become very big which eclipse has a hard time handling (a refresh of the project takes about 5 to10 minutes).
In this particular case, going for a clean OOP solution like I’ve been going for in the past will not give good performance because, as I understand it, it is impossible for the VM to optimize well.
At this point the status of the 68k core is that it runs the 2 game drivers I made (Food Fighter and Snow Bros.) for a few minutes but then goes wrong. Because the problems appear after a few minutes, the bug(s) is/are VERY difficult to find (it appears after billions of correctly emulated instructions) and I’ve been unable to track them down them yet. Maybe I’ll have to implement a driver for a console (maybe Genesis or something) to be able to test it against a large number of games.
Another issue is that it isn’t cycle accurate yet. I simply let each instruction take a fixed number of cycles (a number which seemed like a good estimate), but I don’t think this is a big issue as I suspect it will only affect some games which do some heavy trickery that depend on timing a lot. IIRC, StarScream (another 68k core, but asm based) did basically the same thing but still works for most applications.