68020 AGA core - 2 game issues

[ransom]

Rainbow Islands & Fire & Ice games in whdload do not work is 68020 AGA mode..

Loading error reads

Error: Exception illegal instruction $ 10
at $ 7D7B6

I believe there may need to be a fix made for this as it seems.to be a timing issue.

[mahen]

Hi ! Hmm, I remember playing the AGA version of Fire & Ice with no issue. Which one did you try ? I think the slaves were updated quite recently.

Edit : whdload slave updated several times, incl. in may 2020

[ransom]

Hi ! Hmm, I remember playing the AGA version of Fire & Ice with no issue. Which one did you try ? I think the slaves were updated quite recently.

Edit : whdload slave updated several times, incl. in may 2020

Incorrect, there is no Fire & Ice AGA Slave...

FireAndIce_v3.3_0502.lha - won't work
FireAndIce_v3.2_CD32.lha - works

RainbowIslands_v1.4.1_00112.lha - won't work

1.jpg (1.44 MiB) Viewed 12258 times

2.jpg (1.2 MiB) Viewed 12258 times

3.jpg (1.37 MiB) Viewed 12258 times

4.jpg (1.58 MiB) Viewed 12258 times

[petarku]

just to chime in that on my real amiga 1200 last time when i was running fire and ice whdload didnt work so i was using cd32 version.

[ransom]

just to chime in that on my real amiga 1200 last time when i was running fire and ice whdload didnt work so i was using cd32 version.

Do the two games in question work on your real a1200 amiga?

[shertz]

I tried Rainbow Islands on my Minimig and does not work for me either. I get the same error message.

FYI, I put my minimig into A500 mode and had a floppy image of Rainbow Islands and it worked fine but in Pal mode. It was glitchy in NTSC mode.

[ransom]

I tried Rainbow Islands on my Minimig and does not work for me either. I get the same error message.

FYI, I put my minimig into A500 mode and had a floppy image of Rainbow Islands and it worked fine but in Pal mode. It was glitchy in NTSC mode.

Yes I'm aware a500 floppy versions work in PAL mode..

The two games in question Rainbow Islands and Fire & Ice both work in a1200 AGA emulation mode, which emulates a real a1200 AGA so I dont see why these games won't work on your real a1200.

[petarku]

I remember I had issue with one setup on real amiga 1200 where fire and ice didnt work but cd32 version was working fine.
However i just checked on my amiga 1200 and fire and ice and rainbow islands are working fine so maybe it is just my memory tricking me.

[mahen]

Incorrect, there is no Fire & Ice AGA Slave...

Well, I was referring to the CD32 slave, which has no drawback because it can use ingame modules and not only CDDA tracks.

But indeed there are quite a few OCS games not starting and indicating "illegal instuction". Another example is Iridium and Simulcra. All of them do work in 68k / OCS mode.

[ZigZag]

Incorrect, there is no Fire & Ice AGA Slave...

But indeed there are quite a few OCS games not starting and indicating "illegal instuction". Another example is Iridium and Simulcra. All of them do work in 68k / OCS mode.

I had the same issue with some OCS WHDLoad games on AGA setup using MegaAGS. I'm fairly sure I found a way around it by changing some settings/files in Amiga (I was trying to get "Mercenary" running). Unfortunately I updated MiSTer a week or so ago & it messed everything up so I reinstalled & don't currently have Amiga stuff available (I'm focused on AO486 ATM).

I think if you do your own installation of Workbench & WHDLoad, including the proper ROM files in the DEVS folder the OCS games work just fine on AGA system (my memory could be incorrect, I've been unwell & very tired recently). Perhaps including my own original reg key file for WHDLoad also helped. Once I get back on Amiga stuff I'll reinstall & give more details.

[mahen]

@ZigZag : well, actually most OCS games do work through WHDLoad once properly installed, but once in a while I stumble upon one which doesn't (the ones mentioned above).
Another unrelated issue, compatibility wise, is games have gfx glitches when the D-cache is enabled.

cheers !

[ZigZag]

The Amiga was not just "a computer", Amiga's were lots of different computers. While the AGA ones were generally backwards compatible, most older games required specific settings or software patches. WHDLoad does an incredible job of making all the games work fine on a single Amiga system. It would be surprising if a few incompatibilities/bugs did not continue to arise on any given setup though. I had issues on my real A1200 plenty of times, it's not unusual for things to need tweaked. It's very easy for someone unfamiliar with the hardware & software of the era to mistake a glitch/error that would occur on real hardware for an emulation (or FPGA reimplementation) bug. I'm not suggesting anyone in particular here is unfamiliar. It's not a criticism anyway, unless you grew up using certain machines they can be hard to wrap your head around, I applaud & appreciate everyone who has a love of learning to use old & unfamiliar hardware & OS's. I had an Amiga 1200 since my teens & even after 25 years I still find new ways to hack or trick software into working glitch free on it.

In short, what I'm saying is without testing the same MiSTer setup on the the exact same real HW & OS it's incredibly hard to know if the bug is MiSTer core related or not.

[KnC]

you will find almost all craftgold Games do not work on minimig when using anything other then the 68000 cpu
the 68020 is still not fully implemented/bugged it was traced to VBR but not where or what was causing it as far as i remember

Edit: Graftgold not Craftgold games

[ransom]

Looks like I've figured out the problem myself...

Looks like I've figured out how to play Rainbow Islands & Fire & Ice HDF with Mister....

Make an Amiga 600 Config and save it....

Basically you use a 68000 CPU, AGA, KICK 3.1, then attach the hdf like normal. Both games now work. I've attached screenshots of my a600 config if you wanted to replicate it and save an a600 config..

[mahen]

Yep, AFAIK, the 68000 implementation is cycle accurate contrary to the 68020 one.

@ZigZag : you're right, one has to always double check with a real Amiga or emulated one with the proper settings (same configuration as the MiSTer one) to make sure the crash / bug is due to the Minimig core and not due to the game / whdload slave itself. As stated above, there are incompatibilities due to the 020 implementation.

Let's gather some precise reports for Chaos

[Lisko]

Yep, AFAIK, the 68000 implementation is cycle accurate contrary to the 68020 one.

@ZigZag : you're right, one has to always double check with a real Amiga or emulated one with the proper settings (same configuration as the MiSTer one) to make sure the crash / bug is due to the Minimig core and not due to the game / whdload slave itself. As stated above, there are incompatibilities due to the 020 implementation.

Let's gather some precise reports for Chaos

Yes, AFAIK the 68ec020 isn't fully implemented and some instructions are missing, that should be the reason of getting illegal instruction error.

[lroby74]

Sword of Sodan is another one that launched with 68020 cpu has a weird behavoiur while with 68000 is perfect (on real A1200 is perfect with WHDLOAD and ADF both), hope to see improvements on 020 code very soon

[kolla]

When 68000 is selected as CPU, the MiSTer will use the cycle exact Fx68k core from ljor, while if you select 68020, it will use the tg68.C core by tobiflex. So two different CPU cores. TG68 started out as a 68000 core, and 020 functionality was only added years later. It works amazingly well, but isn’t “perfect”. Tobiflex has mentioned that he has been playing with the idea of making a new 32bit 020 core (named NG68), but it probably will not be ready anytime soon. In the meantime, tg68 will improve

It’s a little funny that one can select 68000 and AGA, as such a system never existed in real life (and I’m unsure if it could have existed.)

[rhester72]

Of course 68000 and AGA could have coexisted - why not? There was simply no reason to - 7MHz 68000 was a joke by the age of AGA. You *could* have 8088 and VGA, too, but why?

[akeley]

When 68000 is selected as CPU, the MiSTer will use the cycle exact Fx68k core from ljor, while if you select 68020, it will use the tg68.C core by tobiflex. So two different CPU cores. TG68 started out as a 68000 core, and 020 functionality was only added years later. It works amazingly well, but isn’t “perfect”.

The number of software titles that require cycle exact 020+AGA is... zero

Changed your mind?

[kolla]

No, why do you ask?

[kolla]

Of course 68000 and AGA could have coexisted - why not?

Because 32bit databus essentially - real AGA uses 32bit transfers for bitplanes etc, while 68000 (and ECS) only does 16bit. That’s why I doubt you can just connect real AGA chipset to a real 68000. IIRC, the Minimig AGA simulates 32bit transfers by doing two 16bit transfers.

[rhester72]

And that's exactly how you'd build the bus multiplexer in real life to do the very same thing, demonstrating quite conclusively that it was certainly possible.

[akeley]

No, why do you ask?

You're a funny guy. Nevermind

Meanwhile, I'll definitely keep on insisting that it'd be great to see A1200 brought as close to 1:1, like I did in this other thread, since it seems there are reasons for doing so after all.

And I'll most definitely take opinions of assorted "experts" with an even bigger pinch of salt than before.

[kolla]

Well, I'm afraid you will never see cycle exact 68EC020 unless you do it yourself.

Do you think those AGA titles above that struggle on tg68 work well on real A4000 with 030 or 040?
That should tell you what you need to know regarding the need for cycle exact 68EC020.

[Chris23235]

Slingshot (Gyurco) over at the Atari Forum has an idea, what the problem might be:

https://www.atari-forum.com/viewtopic.p ... &start=525

[kolla]

I just played both those two games on real Amiga system, but instead of cycle exact 680EC020 A1200, I used one 060 A1200 and one 060 A4000. So again, the problem with these games is NOT that there isn't a cycle exact 68ec020, but simply limitations and bugs with the current TG68 in 020 mode - as usual.

[kolla]

https://www.atari-forum.com/viewtopic.p ... 42#p406042

No “1:1 cycle exact” needed

[akeley]

Marvellous Seems some extra work was needed after all.

[kolla]

Seems some extra work was needed after all.

Nobody suggested that no work is needed, a lot of work is still needed to improve TG68, but "cycle exact 1:1 stock A1200" is NOT a needed goal.