Display Issues When Running Demo Reel 3

[DocFlareon]

Dynamic High-Res screen in the Digi-View portion of the presentation appears garbled with garbage scanlines. Is there a host display option that renders it cleanly?

[Toni Wilen]

Do you mean garbage everywhere or only at the bottom of screen?

If everywhere: make sure config is full cycle-exact 68000 (Normal 0.5M+0.5M A500 config with 2 floppy drives). It still will sometimes have garbage, reason unknown. Possibly bug in demo.

Bottom garbage is demo bug. It is very visible if PAL, less visible (only about 2 lines of garbage) if NTSC.

[Rotareneg]

Tried it on 5.3.0b5 with an NTSC A500 quickstart config and the corruption flickers, some frames are clean and others the whole screen has glitches:



I don't remember that image glitching way back when it was new, but as the .ADF I'm using to test now was found on the EAB file server, who knows what it's history is, like if it's an earlier version or made for PAL only, or whatever.

The same shot from the DigiView 4.0 HiResDemo disk is fine however:

[ross]

Originally Posted by Rotareneg

Tried it on 5.3.0b5 with an NTSC A500 quickstart config and the corruption flickers, some frames are clean and others the whole screen has glitches:

Could you please upload a statefile?
For some reason I can no longer replicate the problem, but it certainly exists and we need to understand why

EDIT: ok, it happened to me too, and I probably figured out the cause.
So don't worry about uploading.
I now have to understand why it doesn't happen on a real machine or perhaps only in exceptional cases.

[ross]

Ok, a statefile that has full screen garbage in PAL is required.
I hope someone can generate it
(in NTSC the problem has been solved)

[Zarnal]

Hello, I made a savestate in PAL and in full screen (with the same floppy disks with a Quickstart A500, I don't know if a specific PAL version exists ). The glitch is at the bottom of the screen.

[Toni Wilen]

Bottom corruption in PAL mode is demo bug, NTSC was slight corruption in last 2 lines only (which was mentioned previously). Happens on real HW too.

[ross]

Just to clarify.

It's unclear if actually the full-screen garbage bug also exists in WinUAE in PAL version (like the one in NTSC in message #3).
[now WinUAE properly support it in NTSC]

It's possible that both Toni and I saw it using KS1.3, which has a bug where the PAL system is randomly recognized as NTSC.
The demo adapts the CPU synchronization code depending on the television system used and is 'wrong' if the KS tells it to be on an NTSC system when in fact it is not.

Therefore it is absolutely possible that the Demo in PAL has always worked well, until proven otherwise, that is, until someone manages to provide a statefile with the garbage in full screen in PAL (and not just in the last lines at the bottom).

[Rotareneg]

Posted this in the 5.3.0 beta thread (with a savestate included,) but I thought I'd mention it here too:

The dynamic high res image is fixed in 5.3.0b6, but now the loading screens are glitchy:

[Waccoon]

Originally Posted by Rotareneg

...but now the loading screens are glitchy:

When looking at the glitching in the visual debugger, I can see that on hpos cycle 0, CPU cycles are able to steal the bitplane DMA on cycle 0, and sometimes I've also seen a refresh cycle conflict that messes up the DMA for the rest of the line.

I have no idea how the "estimated cycles" code works, but this new line in beta 6 seems to be the culprit:

ecycs = maxhposeven ? estimated_cycles_buf0 : estimated_cycles_buf1;

[Toni Wilen]

It is correct but other parts of emulation does not like it. I don't think this situation was ever worked properly (NTSC + last BPL1DAT using cycle 0).