Work on the Minimig core?

[lips2k15]

I was wondering why we havnt seen any work on the minimig core... this could do with some work in many places, RTG and perhaps even a faster cpu mode? I was thinking the RTG could be done with the help of the Vampire guys? Thaughts and suggestions guys?

[kolla]

When Apollo Team is done with SAGA, it is supposed to be open sourced, so why not wait for that. Most useful for me would be faster networking option, or "simply" make some USB port available directly from Minimig.

[lips2k15]

yes I totally agree... is anybody actually working on this?

[sardaukarsiet]

I'm not sure of how much the core _can_ be improved with MiST's older model FPGA. Is RTG even possible?

[slingshot]

I'm not an Amiga expert, but AFAIK RTG is just a common graphics API, where the underlying hardware can be different. So I would say, it's possible to "invent" some capable graphics card on MiST. Just the question is as usual, who has the skills, knowledge and willingness to do it.
However I would be more happy to have a bugfixed CPU first.

[DanyPPC]

I quote:

However I would be more happy to have a bugfixed CPU first.

[MasterOfGizmo]

I have spent quite some time fixing the 68020 instructions and addressing modes. If you can tell me which instructions you think are broken then I might be anle to help.

[DanyPPC]

Pinball Illusions AGA (ADF and WHDLoad version) doesn't work.
Even the last Reshoot-R AGA doesn't start on minimig core.

I think they use 68020 code.

[slingshot]

I have spent quite some time fixing the 68020 instructions and addressing modes. If you can tell me which instructions you think are broken then I might be anle to help.

I can refer to this MikeJ post:
viewtopic.php?f=117&t=34568&p=356031#p356031
CAS2, CMP2

And this bug report:
https://github.com/mist-devel/mist-board/issues/109
MOVEC maybe?

I wonder if there's an 68020 instruction tester for emulators?

[sardaukarsiet]

I also got some errors with that error box on the top left about disk access and continue or something while running OS 3.9 - what should I do when I get them to use as a good bug report?

[slingshot]

First, how can a dev reproduce the bug consistently.
Second (if you want the fix soon) debug what's wrong.

[sardaukarsiet]

I just got errors on bootup after installing 3.9. Is sharing HDFs possible? I can also try running them in WinUAE to see if the bug happens there too.

[slingshot]

I just got errors on bootup after installing 3.9. Is sharing HDFs possible? I can also try running them in WinUAE to see if the bug happens there too.

Of course cross-checking in an emulator is always good (I would say that should be the #1 to check). You can share the HDF, if you want. But currently not many people are working on the core.

[retrofun]

I have done some fixes for the MiST firmware and the minimig-core that improves stability and user experience.

mist-firmware:

• Fix fatal error 4 when choosing new FPGA core
• Make Kickstart 1.2/1.3 honour changes of memory configuration
• Improve stability of core startup when a new configuration is loaded
• Fix IDE on/off, ask to reboot when changed

minimig-mist:

• Disable joystick1/mouse when OSD is active
• Code cleanup with changes from Minimig-AGA_MiSTer

Details & binaries: https://github.com/retrofun/mist-binaries-test

Works fine for me. Please test, comment and (maintainers) cherry-pick!

[slingshot]

Please send a PR to mist-devel/mist-firmware.

[retrofun]

PR created. Fixes for the mist-firmware were pulled and are available with firmware 190820.

[retrofun]

And this bug report:
https://github.com/mist-devel/mist-board/issues/109
MOVEC maybe?

Debugged this. It's the RTD instruction in mmu.library that tg68k doesn't support.

Some days later I discovered in the minimig/MiSTer thread that R4MS already tracked it down to this CPU instruction.

[slingshot]

Now I tried the minimig core again to see Retrofun's patches, and noticed that if I choose higher resolutions, the screen is wrongly blanked (too much border on the left, the main window cut at the right). It's not because of overscan setting.
Then I checked Amiga500 schematics, and realized that:
- no blanking output from Agnus
- csync is the blanking input for Denise
- in the core, the blanking registers are not affecting bitplane fetching
So the question is: how the blanking registers are used in OCS/ECS? (HBSTRT/HBSTOP/VBSTRT/VBSTP)

[Higgy]

@Slingshot are you on EAB ? If not I can send a PM to Toni Wilen, he created WinUAE THE Amiga emulator. I expect he knows the answer. I don't 'know' him but he has been very helpful in other people's projects.

[slingshot]

@Slingshot are you on EAB ?

What is EAB?

[DrOG]

@Slingshot are you on EAB ?

What is EAB?

English Amiga Board?

[slingshot]

@Slingshot are you on EAB ?

What is EAB?

English Amiga Board?

Oh, plausible.
Btw, only me who experienced this blanking issue? Just choose the multisync monitor type, select a 640x480 resolution, and watch the cut down right side of the main window, and the huge border on the left side.

[Higgy]

I was going to spell out EAB, but I thought if you did not know, then you do not go there.
I thought Multisync support was only with AGA chipset (A1200 etc). Back in the day I had a multisync monitor with my A1200. I have feeling that some of those resolutions were not simply 15Hz x 2.

Here is some info:

https://people.debian.org/~cts/debian-m ... cs_aga.txt

Which Monitor Type are you selecting?

[slingshot]

I was going to spell out EAB, but I thought if you did not know, then you do not go there.
I thought Multisync support was only with AGA chipset (A1200 etc). Back in the day I had a multisync monitor with my A1200. I have feeling that some of those resolutions were not simply 15Hz x 2.

Here is some info:

https://people.debian.org/~cts/debian-m ... cs_aga.txt

Which Monitor Type are you selecting?

It doesn't matter, multisync or scandoubled pal, blanking is wrong. I think the default video modes are using the built-in blank values, so they're right, because hand-calibrated by the core designer, but these are using the programmed blanking, and the programmed values are not working well with the current blanking code. So would be good to know how it's really working, done by Agnus or Denise, which counters it is using (I can experiment until it works, but it's a bit exhaustive to do this trial and error job again, would be better if somebody already knows the answer).

[vebxenon]

Same happens here as Slingshot has told.