Problematic Demos.

[Toni Wilen]

Should be fixed now. In theory this could also have happened on real hardware but possibly only after few dozens or hundreds or even more retries..

This again changes corruption in other demos (which isn't important, corruption is corruption). This is low priority feature that won't be fully examined and implemented until later versions.

[amilo3438]

First Demo (Demo) by Starline --> Seems to be fixed with today's winuae.7z!

La Weird (Demo) by Cave --> Still stops with blank screen!

(tested config: Quickstart A500 1.3 ECS 512+512)


EDIT: Music Demo 1 (Demo) by Keo & Mad Max (Jochen Hippel) & Phantom & Tony

Checking the music on this demo (specially F2) with today's winuae.7z my sensitive ears feels degradation in sound quality in comparison to yesterday's winuae.7z or b31. (sorry but have to complain)

[Toni Wilen]

Originally Posted by amilo3438

First Demo (Demo) by Starline --> Seems to be fixed with today's winuae.7z!

La Weird (Demo) by Cave --> Still stops with blank screen!

(tested config: Quickstart A500 1.3 ECS 512+512)

I never said anything else was affected. It does not even have any conflict glitches..

EDIT: Music Demo 1 (Demo) by Keo & Mad Max (Jochen Hippel) & Phantom & Tony

Checking the music on this demo (specially F2) with today's winuae.7z my sensitive ears feels degradation in sound quality in comparison to yesterday's winuae.7z or b31. (sorry but have to complain)

I don't have sensitive ears = need some scientific details. You sure it isn't just some random slowdown or hidden log flood etc? (and always confirm using some previous version because it could be just something outside of winuae)

EDIT: Any sound related reply gets this kind of autoreply because it is almost NEVER anything to do with sound. Thats why real reason needs to be searched, at least basics.

EDIT2: redownload, some blitter test was left in previous update but I don't see how it can affect anything.

[amilo3438]

Originally Posted by Toni Wilen

I don't have sensitive ears = need some scientific details. You sure it isn't just some random slowdown or hidden log flood etc? (and always confirm using some previous version because it could be just something outside of winuae)

EDIT: Any sound related reply gets this kind of autoreply because it is almost NEVER anything to do with sound. Thats why real reason needs to be searched, at least basics.

EDIT2: redownload, some blitter test was left in previous update but I don't see how it can affect anything.

Forget it! (many will not realize it anyway)

How I can explain it... well, b31 till yesterday's winuae.7z has clear interesting to listen sound, and today's winuae.7z makes my ears tired. (this is the best how I can explained it)

EDIT: Did a quick test with latest winuae.7z (afternoon) version and it seems to fixed it! (maybe need more testing but seems ok for now)

[Zarnal]

Originally Posted by amilo3438

-----------------------------------------------------------------------------------------------------------------------

Dots (Demo) by Phenomena

[A500 1.2 OCS 512] some glitches appears on the upper half of the screen!? (same on 4400!) EDIT: Also glitches in A500 1.3 ECS 512+512 but different!

-----------------------------------------------------------------------------------------------------------------------

No glitches on real A500 ( pic 1 ).

[amilo3438]

@Zarnal
Thanks! (I frogot that this one was not checked)

[Toni Wilen]

I am going to delay that one because it needs even more detailed debugging.. (which I did years ago but don't remember exact details anymore)

La'Weird should work again.

[Toni Wilen]

Originally Posted by amilo3438

Dots (Demo) by Phenomena

[A500 1.2 OCS 512] some glitches appears on the upper half of the screen!? (same on 4400!) EDIT: Also glitches in A500 1.3 ECS 512+512 but different!

A500 confirmed. It looks like it shows uninitialized memory which depends on config etc. Possibly it works if run from "original" release (packdisk, whatever)?

[Zarnal]

Originally Posted by Toni Wilen

A500 confirmed. It looks like it shows uninitialized memory which depends on config etc. Possibly it works if run from "original" release (packdisk, whatever)?

I don't really understand anymore .


The demo was copied to a blank ADF.

The test was done on an A500-512ko-1.3 / OCS ( with video ).




With same ADF:


WinUae A500-OCS-512k-1.3: glitches, real A500 stock-512k no glitches.

[amilo3438]

Originally Posted by Toni Wilen

La'Weird should work again.

Confirmed to work in todays (evening) winuae.7z version!


Dots (Demo) by Phenomena

Originally Posted by Toni Wilen

A500 confirmed. It looks like it shows uninitialized memory which depends on config etc. Possibly it works if run from "original" release (packdisk, whatever)?

No, it behaves the same! (tested on "Compact Disk 19 (Packdisk) by Humanoids" with Qs A500 1.2 512 config.)

Originally Posted by Zarnal

I don't really understand anymore .

The demo was copied to a blank ADF.

The test was done on an A500-512ko-1.3 / OCS ( with video ).


With same ADF:

WinUae A500-OCS-512k-1.3: glitches, real A500 stock-512k no glitches.

Interesting that on your test there is no glitches on a real A500!?

[Zarnal]

Originally Posted by amilo3438

Confirmed to work in todays (evening) winuae.7z version!


Dots (Demo) by Phenomena


Interesting that on your test there is no glitches on a real A500!?

I asked the A500 tester for additional information :

A500 used is rev5 with real floppy drive ( no gotek or similar ). No expansions.


Agnus 8371-512ko.




Note : Video available on demand for those interested.

[Toni Wilen]

Finally tried it on OCS A500 and it works without garbage on top of scroller. Only if KS 1.3. If KS 1.2: CLI flash at the bottom happens few times.

This is quite badly coded and seems like was fixed in a hurry, just enough to get it working, mostly..?

It updates copper wait vertical position very soon in vblank interrupt. Unfortunately copper also reads this instruction almost at the same time and if copper reads between CPU writes, it gets too too early position, wait finishes far too early: glitches.

0004AC82 13d8 0004 a200           MOVE.B (A0)+,$0004a200
0004AC88 0639 0080 0004 a200      ADD.B #$80,$0004a200

EDIT: I suspect this was originally coded on A1000 which has delayed vblank interrupt by 1 line, which would completely "fix" this problem.

It also sets one of scroller bitplane pointers to almost end of 1M chip RAM (and modulo equaling negative length of bitplane width). Which equals end of 512k chip ram if Agnus is OCS (this area is zeroed). But it becomes real end of 1M chip if ECS (even if system only has 512k chip). Which is normally not zeroed. Result is tall bars on top of scroller.

And finally, it also pointlessly calls KS vertb handler after it has finished its own vertb stuff. If it happens just in wrong position, systems copper list 2 gets executed at the bottom of screen (CLI window flashes).

More testing needed for correct fix that does not break anything else.

[Zarnal]

Ive tested phenomena-Dots with b31 A1000-512-1.1/1.3 :

Only Cli " glitch " present.


Note : the two others demos ( Cave-8th Union and Circle Scroller by United Forces , with " strange mysterious bug " in Quickstart A500 mode are Ok with A1000-512-1.3 ).

And are Ok with custom Quickstart A500 mode ( advanced chipset 8163/8167 Agnus option ).

[Toni Wilen]

Phenomena-Dot (aka Phenomena-SpectreParty) should work now without scroller glitches.

Rare A1000 CLI flashes is normal.

[Zarnal]

Originally Posted by Toni Wilen

Phenomena-Dot (aka Phenomena-SpectreParty) should work now without scroller glitches.

Rare A1000 CLI flashes is normal.

Confirmed.


With 20H16 version, no scroller glitches. in OCS.

Edit 07/09/2021 : and ECS have normal glitch.


And Cave-8th Union is now Ok ( at least in OCS ).

Edit 07/09/2021 : and in ECS too.


Circle Scroller by United Forces has still the same problem ( just for info ).

Edit 07/09/2021 : OCS/ECS.

[Toni Wilen]

Circle Scroller by United Forces is fixed. Blitter busy bit in DMACONR has 1 cycle delay compared to copper noticing blitter finished state. Blitter busy bit in DMACONR is a flipflop and loading it takes 1 cycle.

Previous WinUAE version used correct DMACONR bit reading timing, then 4.5+ custom chipset updated copper timing (which broke DMACONR reading). Now both should be correct.



Background info: Coder got very lucky..

This demo calls original KS level 3 interrupt handler like many early demos. It also didn't disable level 6 interrupt (CIA-B)

When it calls KS level 3 interrupt handler near end the frame, it checks some variables, then points COP2LC to system second copper list. Then it does the critical bit: it reads current vertical position and writes it to CIA-B TOD low byte.

Bad things happen if current vertical position is zero, it looks safe but because it only reads low 8 bits of vertical position, line 256 equals line zero.

When zero is written to CIA-B TOD low byte, it immediately causes TOD ALARM interrupt (interrupt is enabled and ALARM=000000).

Level 6 interrupt handles runs, CIA-B handler gets confused and assumes blitter queue is running (or something like that), enables blitter interrupt ($8040 -> INTENA). (KS2.0+ does not do this anymore, I think it was a bug that ALARM was left to default zero)

Blitter interrupt is already active (because demo uses blitter heavily), blitter interrupt executes, demo thinks it is vertical blank interrupt and timing goes very wrong (You can notice music slightly speeding up when things start to go wrong)

But because coder got lucky, current line is never 256 when system interrupt handler runs.. It jumps above and below 256 but it is never exactly 256!

Thanks for finding this demo

[Zarnal]

Originally Posted by Toni Wilen

Thanks for finding this demo

It was Amilo3438 who found this demo (and all the others). I did nothing but retest behind him (and "my" A500 tester on the AF forum did some verifications on the real hardware before making a report, Phenomena-Dot (aka Phenomena-SpectreParty) for example and others) .


Usually I only test AGA but at the moment I also test OCS (and "my" tester checks first ).


Amilo3438 does an extraordinary and exemplary job. And I know the time and focus and patience it takes.


And confirmed for Circle Scroller by United Forces. It's now Ok.

[amilo3438]

Originally Posted by Toni Wilen

Background info: Coder got very lucky..

I was lucky too finding this demo!

Great that you manage to fix it!

Originally Posted by Zarnal

It was Amilo3438 who found this demo (and all the others). I did nothing but retest behind him...

You have also contributed in your own way!

[Zarnal]

Originally Posted by amilo3438

Boze and Bobs (Demo) by Area 08

[A500 1.3 ECS 512+512] bobs glitches!? (fine in 4400!)

(tested in 4900b30 and latest winuae.7z)

Do you have any new information regarding the cause of the problem with this demo ?

[amilo3438]

Tested in b32 or latest winuae.7z
-----------------------------------------------------------------------------------------------------------------------

All Those Years (Demo) by Phenomena

[A500 1.3 ECS 512+512] In ECS the music slowing down when corruptions appears! (in OCS is fine as expected)

(Ps. I reporting it mostly because of the music is slowing down in ECS so not sure does it happens on a real machine.)

-----------------------------------------------------------------------------------------------------------------------

Anzax First (Demo) by The Lost Boys

[A500 1.3 ECS 512+512] If ECS there are glitches on he left side! (in 4400 it is different)

-----------------------------------------------------------------------------------------------------------------------

Capone (Import Intro) by Bitstoppers

[A500 1.2/1.3 OCS/ECS 512+512] On the bottom scroller some white points appear! (if CE=disabled seems fine)

(Ps. If CPU=68010 white points does not appear but bottom red stripe glitches?)

-----------------------------------------------------------------------------------------------------------------------