Game Glitches on MegaAGS

[Malor]

It ran fine the last time I tried it. I never get very far with it, but the sound and graphics, at least at the beginning, seemed flawless as of 60 or 90 days ago.

[vanfanel]

The problem seems to be in the 68020 CPU implementation.
Running the game (WHDLoad install) in plain 68000 mode, glitches don't happen. They only happen when running it with the 68020 CPU: DCache or Chipset settings don't have any impact on the problem.

And it's not MegaAGS either: I use my own system installation and the same problem is present with 68020 CPU.

[msimplay]

I am having severe graphical corruption in Street Fighter II using Mega AGS but I am guessing it's in the Minimig core itself as I don't have this issue on Winuae or an original Amiga 1200

[wark91]

Have you d-cache set to ON. Try to disable it. It helps some games when it is off.

[msimplay]

The dcache is off I have tested both on and off and it made little difference.
I am hoping someone else could test the game I know the game is not good on the Amiga however it should be working correctly?

[fierman]

The mega-ags setup does not like zorro3 ram. Try setting fastram to a different value. The game runs fine here, so it is definitely not a minimig issue

[msimplay]

The mega-ags setup does not like zorro3 ram. Try setting fastram to a different value. The game runs fine here, so it is definitely not a minimig issue

What's the value you have it set to?

[fierman]

The mega-ags setup does not like zorro3 ram. Try setting fastram to a different value. The game runs fine here, so it is definitely not a minimig issue

What's the value you have it set to?

hum. I see the glitch now too. I swear it was not there before. (in any case - 8MB fastram should be more than enough)

[msimplay]

Might be something that has happened in recent updates

Edit:

I tried lowering the ram to 8mb last night and made no difference to the graphics corruption in that game sadly.

Also I did try out Mortal Kombat 2 on the off chance and that also seems to have some issues with sound.
I was playing through the Arcade mode and after some time the sound begins to distort and buzz

[Juri]

have you tried OCS or ECS instead of AGA?

[msimplay]

have you tried OCS or ECS instead of AGA?

Yes that didn't do much.
I have also tried to change the CPU to 68000 instead of 68020 but that stopped MegaAGS booting.

Have you tested the game to confirm it works without corruption on your setup?

[Juri]

Have you tested the game to confirm it works without corruption on your setup?

i do not like shadow dancer, but i tested it. yes i have glitches when jump. i tested the game in 68000 mode, ECS, 1Mb chip and 8MB fast ram (no slow ram)
glitches gone away.

[msimplay]

Have you tested the game to confirm it works without corruption on your setup?

i do not like shadow dancer, but i tested it. yes i have glitches when jump. i tested the game in 68000 mode, ECS, 1Mb chip and 8MB fast ram (no slow ram)
glitches gone away.

Shadow dancer works but in my previous posts I was talking about Street Fighter two specifically I posted a video to demonstrate the issue also

[Juri]

can't find the video, sorry i'm blind and without goggles.
tested and i do not have any glitches. but i do not use mega ags, i have a self made whdload collection. 68020, 2mb chip, 256mb fast, 0mb slow, AGA and kick3.1.4

[msimplay]

can't find the video, sorry i'm blind and without goggles.
tested and i do not have any glitches. but i do not use mega ags, i have a self made whdload collection. 68020, 2mb chip, 256mb fast, 0mb slow, AGA and kick3.1.4

Please try Street Fighter 2 and let me know if that works for you thanks

Also here is the video for reference

https://www.dropbox.com/s/r5unvbjxzj02p ... 2.mp4?dl=0

[Juri]

yes it work for me in my whdload collection
i will try the mega ags collection one in weekend

[Optiroc]

Apparently the Street Fighter 2 games are set up to boot into NTSC mode, and I seem to remember that made them slightly more bearable and, more importantly, that used to work without introducing glitches. But that was when MegaAGS was first conceived and much have happened to the core since then. I'll look into a fix.

Edit: On my setup Street Fighter 2 is equally glitchy when setting it to load in PAL mode. No idea why it seems to work on other setups, and while it's no loss at all for this game figuring out the root cause may solve other glitchy titles.

[msimplay]

An update I tested with an older image tutorial I found on Youtube to quickly set up Minimig in MiSTer
I found that Street Fighter 2 is working perfectly in this image by someone called Pez

https://youtu.be/VhUsCgTI0o0

EDIT: On the MegaAGS image I changed the chipset to ECS and that stopped megaAGS from starting up however it booted to the desktop.
When I manually clicked on the Street FIghter 2 WHDLOAD image it works without glitches.

There is definitely some kind of misconfiguration in MegaAGS which is causing the glitches in some games

[Optiroc]

An update I tested with an older image tutorial I found on Youtube to quickly set up Minimig in MiSTer
I found that Street Fighter 2 is working perfectly in this image by someone called Pez

https://youtu.be/VhUsCgTI0o0

Interesting find. So something changed between then and now that makes certain things glitch out a bit… Could be something added to the startup sequence, or some changed library. I’ll try to dig into it a bit.

[Optiroc]

I did some testing with old and current versions of the core, and old and current versions of the MegaAGS image. All combinations give the same glitches on Shadow of the Beast and Street Fighter II. These games used to be glitch free at some not-too-distant point in time. I'm certain I tested Shadow of the Beast early this year for example.

So this would point to that the glitches are not caused by changes in the core, nor changes in the MegaAGS configuration. I don't have any good hunches what could've caused this to happen though. Could changes in Main during this year have caused higher pressure on the DDR, which in turn could impact FastRAM timings?

Also, booting in the franked-amiga setup of 68020, lots of RAM but ECS chipset does indeed clear up the glitches. Which isn't that strange by itself since AGA compatibility is not as good as OCS/ECS compatibility, but it does not explain a regression.

[msimplay]

It could be the way workbench is set up.
I noticed that the older image has a more typical workbench set up.
Possibly using classicwb https://classicwb.abime.net/

Where the latest MegaAGS workbench seems more customised when I press escape to get back to the Desktop

[Optiroc]

The plot thickens... So these games (Street Fighter 2, Shadow of the Beast) aren't glitchy on that old TinyLauncher image that started to float around in autumn 2018 (if memory serves), using the same exact core settings that do produce glitches with MegaAGS?

IIRC that's a ClassicWB setup that uses TinyLauncher, but that in itself doesn't point to a root cause here. The MegaAGS workbench is actually much closer to "stock" than ClassicWB.

It's worth noting that the MegaAGS setup is glitch free on UAE and a real Amiga, so I would say this is a core issue with some hard to pin down workaround. I'll do some more experimenting...

[msimplay]

I'll try MegaAGS on my Real Amiga when I get a new CF Card

EDIT:

Lets get a list of all the glitchy games and test them on the old hard drive image.

So far its Shadow Dancer, Shadow of the beast,
Street fighter 2.

In the video it mentioned some games may not work so I'm not sure if all games were working fine there.

[Optiroc]

I tested SotB and SF2 on the very old "AGS2.hdf" image, and both are indeed glitch free. The next step is to install the latest WHDL versions on that image and see if there's anything about the WHDL binary or updated slaves that causes the glitches. If not, it can only be something in the WB install; setpatch version, libraries, and such.

[Optiroc]

I didn't need to go very far down that TODO list to reach a conclusion.

I based a test off the "AGS2.hdf" image mentioned in that Pezz video (a quite stock install that could load both Shadow of the Beast and SF2 without glitches), but copied over the current installs of said games. Et voilà, the same glitches as we get in MegaAGS are there too. The version of the WHDL binary didn't matter; both games could boot with both the latest version and the old one in the original image.

So the glitches have nothing to do with how the MegaAGS system is set up, they are simply introduced by the updated WHDL installs ("slaves") for these games. It's curious that updated version are seemingly less compatible, but I would rather say it underlines the fact that the Minimig AGA implementation has some accuracy flaws.

Workaround: If someone really want to play these games, there are lots of options. One being to boot into MegaAGS in ECS mode and find your way to load into the game using shell commands or the workbench.

[Malor]

Isn't the MegaAGS Kickstart a custom patch? That might be related.

[Optiroc]

Isn't the MegaAGS Kickstart a custom patch? That might be related.

It was experimented with at one point, but now it's the standard 3.1 kickstart.

I checked if the old installs from "AGS2.hdf" are glitch free on the current MegaAGS setup, and indeed they are. That said, I don't see it worthwhile to test all WHDL updates for regressions and if so pin to older versions; we should rely on that the installs and the minimig core are converging into the best way to play these games. There will be some bumps on the road, and that's fine.

[msimplay]

I didn't need to go very far down that TODO list to reach a conclusion.

I based a test off the "AGS2.hdf" image mentioned in that Pezz video (a quite stock install that could load both Shadow of the Beast and SF2 without glitches), but copied over the current installs of said games. Et voilà, the same glitches as we get in MegaAGS are there too. The version of the WHDL binary didn't matter; both games could boot with both the latest version and the old one in the original image.

So the glitches have nothing to do with how the MegaAGS system is set up, they are simply introduced by the updated WHDL installs ("slaves") for these games. It's curious that updated version are seemingly less compatible, but I would rather say it underlines the fact that the Minimig AGA implementation has some accuracy flaws.

Workaround: If someone really want to play these games, there are lots of options. One being to boot into MegaAGS in ECS mode and find your way to load into the game using shell commands or the workbench.

Yes I can confirm I transferred the latest WHDL version of Street Fighter 2 because it has cheats for unlimited credits and yes this is the version that introduces the glitches.

While it's not ideal I guess we should use the older versions of the WHDL files as a workaround or even add both versions if that's possible until the core is updated to be more accurate to the OG hardware as of course these bugs are not present on the OG hardware

My guess is that the trainer functions introduce the bugs

[Optiroc]

Yeah.. And again, it’s important to note that the bugs are in the core, not in the WHDL installs.

[Malor]

Do the same bugs show up on the MiST version of Minimig?