Problematic AGA demos.

[paraj]

Originally Posted by Amcio

Yes, It does! Thanks Alot!

Great! I just tried altering various settings between your config (which also glitched for me) and mine until it worked.

The intro must do something silly since it also glitches on my pistorm32 when run normally, but works w/ a generic wb31+fpu whdload (well it glitches a bit, but not w/ nocache).

Almost curious what it does, but don't recognize "MWC\x01" packer, so don't think I can be bothered.

[Don_Adan]

I dont see depacker code, but MWC01 can be fake header for RNC01 (ProPack).

[BackSpace]

Hello to all,

i know that these two 64k demos problems have already been quoted but i've noticed an evolution of their issues (from no major issue to sound and Gfx issues).
The demos are Zeon and Grid2 both from Nature Group.

The issues are mainly trashed sound.
Discussions about Zeon are here and the workaround is here.
This workaround and all suggested option did not work for me so i've tested older WinUAE version and here what I've noticed

First of all, i've always used QS A1200 Blizzard 1260 configuration for each test.
I've used a blank HDD and i've installed Workbench 3.1 from scratch and have copied these Libs for the Blizzard support.
So i've Workbench 40.42 with Kickstart 39.106

- For Zeon i've always run Nature-Zeon from Aminet (crc32 842D68F3)

- With WinUAE 3.6.1 i've set Fastest possible option for smooth execution (JIT is Off) and Zeon have no issues at all
- From WinUAE 4.0.0 to 4.9.1 i've set Fastest possible option for smooth execution (JIT is Off) and Zeon have sound issues mainly at the beginning but also at the onescreen credit part
- With WinUAE 5.0.0 i've set Fastest possible option and for smooth execution
if JIT is Off the loading music is trashed and there's noise whole demo
if JIT is On there's same sound issues from 4.x WinUAE branch
- With WinUAE 5.1.0 same problems as 5.0.0 but the endpart have gfx issue, some additional stripes have appears hidding the scrolltext and the end of the text to the left is not "dotted" (see captures, text normally appear the the right in diagonal and disappear in dots to the left)
- With latest Beta to date (WinUAE 5.2.0 B4) same problems as 5.1.0

As mentionned the workaround (No flags) and all suggested option (sound buffer, vsync, ...) did not work for me...

- For Grid2 i've always run Nature-Grid2final from Aminet as this is the latest published version (crc32 4A6AB224)

- With WinUAE 3.6.1 i've set Fastest possible option for smooth execution (JIT is Off) and Grid 2 have no issues at all
- From WinUAE 4.0.0 to 4.9.1 i've set Fastest possible option for smooth execution (JIT is Off) and Grid 2 have sound issues start at the wiretunnel part (@1min50sec) and continue with a repetitive noise until almost the end of the demo
- With WinUAE 5.0.0 and > i've set Fastest possible option and for smooth execution
if JIT is Off the music is slowed whole demo
if JIT is On there's same sound issues from 4.x WinUAE branch

As i've used QS configuration i suppose it's not a configuration problem mainly for the Gfx endpart issue of Zeon.

Maybe some options must be set to On/Off or some settings must be modified, have you any ideas?

Thanxx.

[StingRay]

Originally Posted by paraj

Almost curious what it does, but don't recognize "MWC\x01" packer, so don't think I can be bothered.

Executable is crunched with CrunchyDat.

[Zarnal]

Heretia-So
Quickstart A1200

New issue since WinUae 5.1. Logo " Heretia " is corrupted ( pic 1 ). It's ok with WinUae 5.0 ( pic 2 ).

[amilo3438]

Originally Posted by Zarnal

Heretia-So
Quickstart A1200

New issue since WinUae 5.1. Logo " Heretia " is corrupted ( pic 1 ). It's ok with WinUae 5.0 ( pic 2 ).

With 2x (A500) frequency no problem!
So 4x (A1200) frequency seems fast for the demo! (I'm not sure if that falls under the problem!?) EDIT: It will work with disabled CPU caches! (So it could also be cpu cache emulation working too fast as a problem!)

Tested with: Intel Outside 10 (Packdisk) by Ram Jam --> http://janeway.exotica.org.uk/release.php?id=15135 (Note: Demo number 5!)

Ps. Internal timing of 68020 sequencer is still not correct anyway! (I guess, it is still in development phase.) EDIT: And as it seems, it probably never will be! (without docs, which nobody has, it's like science fiction)

[Zarnal]

Originally Posted by amilo3438

With 2x (A500) frequency no problem!
So 4x (A1200) frequency seems fast for the demo! (I'm not sure if that falls under the problem!?) EDIT: It will work with disabled CPU caches! (So it could also be cpu cache emulation working too fast as a problem!)

Tested with: Intel Outside 10 (Packdisk) by Ram Jam --> http://janeway.exotica.org.uk/release.php?id=15135 (Note: Demo number 5!)

Ps. Internal timing of 68020 sequencer is still not correct anyway! (I guess, it is still in development phase.) EDIT: And as it seems, it probably never will be! (without docs, which nobody has, it's like science fiction)

That, I already knew, it was a mistake on my part not to have checked. I have just resumed mass AGA testing with WinUae 5.3, being more methodical than the last time (with WinUae 4.1) where I took the demos/intros at random.

But, remember, not everything is necessarily linked to a CPU problem (almost all cases it's true but there are exceptions), we have already had some examples in the past (Subspace-Ubergirl, Zomo S(p)id, Passion-Nitte for example).

Then there are also the undesirable effects of the current chipset emulation that Toni talked about. I don't know what these side effects are but I suppose they can cause glitches?

[amilo3438]

Originally Posted by Zarnal

Then there are also the undesirable effects of the current chipset emulation that Toni talked about. I don't know what these side effects are but I suppose they can cause glitches?

Of course, there are other factors that can affect the correct behavior of an emulated machine like the one you mentioned.

But as we know, ECS is much more advanced emulated and tested more with 68000 CPU, so testing demos that requires ECS with 68020 CPU would be ideal in terms of testing the CPU itself. (because it removes those other factors)

I tried to look for them a bit, but I didn't really find much, although there are probably many more.


Here's a short list of some that I've found to work on such a configuration:

BD-LetsTwistAgain
Complex-Wormhole
Moranet-Coredump
MrDeath-MusicFromTheCrypt
PS-FakeOne
Void-XmasLights


The demos above will run fine on a WinUAE Quickstart A500+ configuration with some changes: 68020 CPU, 2Mb Chip RAM, 8Mb Fast RAM

I would like to expand the list, so I would ask if anyone knows of another similar demo?


EDIT: Above demos are now in TheZone!

PS.
I found another one, RamJam-OldStyle, but it looks like the .adf file is broken! http://janeway.exotica.org.uk/release.php?id=21669 Edit: No, it is for AGA anyway!


Ok, found another one:
(Although it is for AGA, it also works on non-AGA!)

Roots 2.0 by Sanity -> https://www.pouet.net/prod.php?which=3251
Space Lamer by I486 -> https://demozoo.org/productions/32976/
Pink Project by Equinox -> http://janeway.exotica.org.uk/release.php?id=10796
Go With The Flow by Drog -> http://janeway.exotica.org.uk/release.php?id=3763
Rescue 911 by Limbo -> http://janeway.exotica.org.uk/release.php?id=11610
Art of 4k by Dual Crew-Shining & Rave Network Overscan -> http://janeway.exotica.org.uk/release.php?id=18830 (Note: Needs also a FPU!)
Polly 68020 by Enemy -> http://janeway.exotica.org.uk/release.php?id=3598
XFig v1.9d 68020 by Crux -> http://janeway.exotica.org.uk/release.php?id=77116

[Zarnal]

It's very interesting. Thanks for your link.

For the 1200, there are generally good results even if the 68EC020 is not accurate.
I was surprised because I expected worse visual results given the circumstances (Quickstart A1200 mode).
The worst case scenario is that in the same demo something is too fast and something else is too slow. Or there are slowdowns that shouldn't be there.

The simple presence of Fast (with WinUae) can correct certain problems (While the demo works normally on real hardware without Fast). The opposite is also true however (from memory, Freezers-Symptom for example).
I also had a few cases with 3D objects slowing down even though it works fine on standard hardware.

Toni, I have a question: is the FPS report reliable with 3D objects if the PC configuration is too old (CPU overload 150% or more). + Quickstart A1200 (sometimes the colors quickly panic in a few seconds on the same scene (gray, green, blue) or sometimes oscillate quickly between green and gray (or vice versa) ? Thanks.

[Mclane]

Originally Posted by Zarnal

It's very interesting. Thanks for your link.


The simple presence of Fast (with WinUae) can correct certain problems (While the demo works normally on real hardware without Fast).

That is odd, why would WinUAE need FAST ram to unbreak something that works without it on real hardware. This isn't a question to you Zarnal, more, so I can find out why this is the case. IE Is WinUAE not working correctly here.

I know having FAST ram on a 1200 configuration is suggested to make things work better / faster.

Whatever the answer(s), WinUAE is simply stunning..

[amilo3438]

A layman's opinion...

With the fast RAM, I guess, the 68020 it doesn't have to fight with chipset for the cpu time, so that's why programs that slowdowns now runs fine!?

But if it's not behave the same as on a real machine, yes it an emu problem!

[Mclane]

Thanks Amilo, I really hate to critique WinUAE considering the MASSIVE amount of work Toni has put in over the 25 plus years he has updated the Windows version. But, if it really is an issue where WinUAE isn't working like real hardware, it then deserves a good looking at.

Toni is very keen to give exact or as near emulation, so I'm sure he would think the same..

[amilo3438]

But I'm afraid that even he can't do much there without new information about the 68020 processor itself (or something else missing), which is not currently available, to improve it even more!

Yes, I agree, WinUAE is simply amazing!