Problematic AGA demos.

[ross]

Originally Posted by Toni Wilen

Very late answer but it is fixed now

(Borderblank + DDFSTOP ending before DIWSTOP needs special handling, normally border after end of DDFSTOP is assumed to be same as real border. Which isn't true if borderblank. I'd have expected this to be visible more commonly..)

Let's see if I understood..
If BRDRBLNK=1 between DDFSTOP and DIWSTOP you a have a "COLOR0 border", after DIWSTOP a "BLANK".
In "BLANK" zone all is invisible and BLACK (including sprites), but in "COLOR0 border" sprites are visible if BRDSPRT is set.

So the error before was an early "BLANK" positon (where also a theoretically covered sprite at the right border is visible).

[Toni Wilen]

Originally Posted by ross

Let's see if I understood..
If BRDRBLNK=1 between DDFSTOP and DIWSTOP you a have a "COLOR0 border", after DIWSTOP a "BLANK".
In "BLANK" zone all is invisible and BLACK (including sprites), but in "COLOR0 border" sprites are visible if BRDSPRT is set.

So the error before was an early "BLANK" positon (where also a theoretically covered sprite at the right border is visible).

It has nothing to do with sprites (sprites are never visible in border if brdblank is set, even if brdsprt is set)

Normally horizontally end of bitplane (this is not DDFSTOP but first pixel after last bitplane pixel) to end of display is always drawn using color 0. This is simple optimization, can draw it as single "line". (and normally last bitplane pixel is outside of display window so this "gap" does not even exist)

Border blanking changes this, end of bitplane to end of display window is color 0 but actual border is black. This was not handled. BPLCON4 non-zero XOR value can also change color of "empty" display window compared to actual border. This was already handled.

[ross]

Originally Posted by Toni Wilen

It has nothing to do with sprites (sprites are never visible in border if brdblank is set, even if brdsprt is set)

Yep, right.

Normally horizontally end of bitplane (this is not DDFSTOP but first pixel after last bitplane pixel) to end of display is always drawn using color 0. This is simple optimization, can draw it as single "line". (and normally last bitplane pixel is outside of display window so this "gap" does not even exist)

Border blanking changes this, end of bitplane to end of display window is color 0 but actual border is black. This was not handled.

All clear, thanks!

BPLCON4 non-zero XOR value can also change color of "empty" display window compared to actual border. This was already handled.

I had noticed it doing some tests

[Zarnal]

Thank you for your fix Toni.

I have found a strange issue with " Tek/Rampage " on A1200 emulation.

Strangely, this demo is ok on circle scene with A500/600 emulation (pic1)

On a A1200 emulation, glitchs appears on circles (pic2)

Not confirmed on real A1200.

[Toni Wilen]

I am not interested and I am quite sure this thread is about AGA demos not working on hardware it was designed to run.

[heavy]

phew what a slap
this demo run fine on a hard stock Amiga 1200. so there is an issue but not an AGA issue.

addendum just for information :
this demo run fine since winuae 1.6.2 beta 4.
That "glitch" appear with version 2.3.1.

[Toni Wilen]

Perhaps but check first post. AGA demos only. Only reason this thread exists because scope is very restricted: AGA demos that run on unexpanded A1200. If it is CPU timing related: I am not interested because there is no way to fix it!

[heavy]

I'm not blind, and I said "it's not an AGA issue"
just be a little bit kind sometime is quite cool

perhaps issue could be reported to the thread
http://eab.abime.net/showthread.php?t=46485

the source code of this part is on github. if some want to take a look
https://github.com/tcurdt/rampage

[Toni Wilen]

1) this is only about AGA
2) this is guaranteed CPU speed problem and there is currently no way to fix it, without someone finding out how 68020 works internally. (No, existing documentation is useless)
3) I ignore anything that is obviously CPU speed related. It is obvious if it works in 68000 AGA mode
4) There is no guarantee that this isn't random or that there are tiny glitches here and there that can be missed too easily.
5) I am not going to bother with being kind or cool. I always go straight to the point. Like it or not.

And finally, there are probably hundreds of demos that accidentally work in unexpanded A1200 config. This kind of compatibility testing is far too boring.

But if you (or anyone else) want to do this kind of testing (with real A1200 confirmation), you can start new thread. I probably ignore it today but it may be useful in the future. Probably.

[heavy]

it's exactly what I suggested : if a moderator could create a new thread from the post of Zarnal, it will be more easier to find it one day.

[Foebane]

I've been having problems with this prod ever since 4.1.0 (I believe it works fine in 4.0.1, but that's no longer available, as far as I know).

The demo plays fine, up until the rotating object just before the end credits and interview (see video). In 4.1.0, that part garbles up nastily with a lot of corruption and affects the palette colours in the end.

[DamienD]

Originally Posted by Foebane

I believe it works fine in 4.0.1, but that's no longer available, as far as I know

http://www.winuae.net/download-old-versions/

Any versions not hosted above, then go to: https://sourceforge.net/projects/winuae/files/WinUAE/

[Foebane]

Originally Posted by DamienD

http://www.winuae.net/download-old-versions/

Any versions not hosted above, then go to: https://sourceforge.net/projects/winuae/files/WinUAE/

Nope, just tried 4.0.1 with the above demo, and it still glitched at the same part, although not as bad as 4.1.0.

The weird part is is that it used to play fine all the way through, and only recently has the glitch occurred. What's more, sometimes it will play fine, but glitch at other times.

I use my own configurations for demos, but I sometimes wonder if the Quickstart Compatibility slider will affect behaviour and override those parts of the configs?

Can you try this demo on your own system, DamienD? Attached is the config I use for it (where it sometimes runs fine and mostly doesn't).

[DamienD]

Originally Posted by Foebane

Nope, just tried 4.0.1 with the above demo, and it still glitched at the same part, although not as bad as 4.1.0.

I just provided you with a place to get "4.0.1" from; as you said it works fine in this version and you couldn't locate it...

Originally Posted by Foebane

Can you try this demo on your own system, DamienD? Attached is the config I use for it (where it sometimes runs fine and mostly doesn't).

Sorry, I don't do demos or really know where to get them from...

[Foebane]

Originally Posted by DamienD

I just provided you with a place to get "4.0.1" from; as you said it works fine in this version and you couldn't locate it...

I downloaded it from the second link you provided.

When I used 4.0.1 last year, that demo ran fine, but even on 4.0.1 it glitched when I ran it today.

I'm just wondering if something other than the configuration is influencing the behaviour of program execution on WinUAE.

[Toni Wilen]

Works fine here (boring plain 68040 config with fast cpu with lots of ram, seems to work even with JIT enabled).

[Foebane]

I just tried running it with No Startup-Sequence, and it worked!

Normally, I run them from a window filled with demo files, but it seems this example needs minimal stuff loaded before it will work.

Just out of curiosity: How does one run a demo archive like that without a system HDF? Drag'n'drop into the purple tick/floppy screen?

[Zarnal]

Originally Posted by SaphirJD

http://www.pouet.net/prod.php?which=17868

Has a glitch - when at minute 1:51 seen here - [ Show youtube player ] - a ball of light leads towards the next scene with the logo, Winuae shows only a glitch instead of the effect.

Used configs: Quickstart A1200, Quickstart A4000 config.

Tested Winuae versions: 3.4.0, 3.5.0 and recent Betas.

Just for info : She stops just before this scene on real hardware (A1200 stock+4Mo).

[Foebane]

Originally Posted by SaphirJD

http://www.pouet.net/prod.php?which=17868

Has a glitch - when at minute 1:51 seen here - [ Show youtube player ] - a ball of light leads towards the next scene with the logo, Winuae shows only a glitch instead of the effect.

Used configs: Quickstart A1200, Quickstart A4000 config.

Tested Winuae versions: 3.4.0, 3.5.0 and recent Betas.

The problems with this prod are NOT down to WinUAE emulation, but a corrupted archive.

I had the same issue with this prod, so I looked around at other download links, and the one on Demozoo (as of now) works fine with WinUAE:
https://files.scene.org/get/parties/...mo/psp-and.lha

I tried to get Pouet.net to change their "Download" link to this one, but they're convinced there's nothing wrong with the current file at the link. I haven't pressed the issue, but Demozoo is usually a good place to find alternate download links.

Hope this helps, after so long!

[britelite]

Originally Posted by Foebane

I tried to get Pouet.net to change their "Download" link to this one, but they're convinced there's nothing wrong with the current file at the link.

Who exactly are "they"? As in, I tried to see if you've discussed this on Pouet, and couldn't find any instances.

I haven't pressed the issue

Why not? If you're absolutely sure you're right on this one, then the link should definitely be changed.