New accelerator for a Amiga, Atari and others

[terriblefire]

My understanding is that TG68 is missing the bus arb logic. Since the amiga doesnt need it.

[exxos]

Why would TG68 be platform specific ?,.. I don't know the answer, but surely it's going to be a minor implementation detail rather than a design issue. We did embedded boards back in the day based on 68000 and 68332 and I don't remember having problems with shared memory peripherals.

Happy for technical answers.

Because it misses some control pins which are used on the ST but not on the Amiga. It's been talked about countless times with the TG68 etc.

[amimjf]

Ok, my quick Google didn't find it,... Must be a bug in the Atari ! 😉

[utri007]

I don't know, but kipper2k said that there could be versions for a Atari and even Mac.

After all, it is FPGA so multiple cores should be possible.

[terriblefire]

I don't know, but kipper2k said that there could be versions for a Atari and even Mac.

After all, it is FPGA so multiple cores should be possible.

I have code for the bus ARB that i will share with Mike if he wants.

[alexh]

If this is based on the TG68 core.. That won't work on Atari.. So I doubt this one will either. Looks like a Amiga booster not Atari ?

It might not now but it can and will.

I have code for the bus ARB that i will share with Mike if he wants.

He will appreciate that.

I think it needs everything you solved doing the Atari version of TF534/536. A different memory map, bus errors etc.

[terriblefire]

Thinking again about this.... The Amiga cannot work without the sync bus (VPA, VMA, E) because the CIAs use this protocol. So this cannot be missing.

Bus error will be needed by Atari and we never see BERR on the Amiga. We use it only to tell the 030 that the FPU is missing so TG68 will not need it. So this is possibly missing. It should be trivial to add this as its just another exception/trap.

The Amiga 500 also doesnt have/need bus arb to work. It would need it for the A570 type CD Drive expansions though. The Atari needs bus arb working to use the floppy drive. So it needs to be perfect to work in that machine. You can get away with it being a bit rough in the A500.

The last thing that may be missing is vectored interrupts. The amiga does "use" these with a real 68000 by placing the vectors at the very end of rom. This is a performance trick because autovectors are so slow on the 68000. But functionally they are the same as autovectors. On the CD32 I bypass these and use auto-vectors because they're faster. However the Atari ST needs these and uses them from the MFP. I'm not saying they are missing just that an FPGA core designer could shortcut these and the Amiga would still work.

EDIT: The vectored interrupt performance vs autovectors is the reason why these bytes exist at the end of rom

Code: Select all

0007FFF0   00 18 00 19  ....
0007FFF4   00 1A 00 1B  ....
0007FFF8   00 1C 00 1D  ....
0007FFFC   00 1E 00 1F  ....

The rom is decoded during an interrupt ack cycle and these vectors are placed on the bus by virtue of the address bus having the interrupt being acknowledged on the lower address pins. The rest of the address bus is high during this cycle. This gives us 18 for interrupt level 1, 19 for interrupt level 2 etc. These correspond to the auto-vectored interrupts.