English Amiga Board

English Amiga Board (http://eab.abime.net/index.php)
-   support.WinUAE (http://eab.abime.net/forumdisplay.php?f=5)
-   -   WinUAE 4.9.0 beta series (Was 4.5.0) (http://eab.abime.net/showthread.php?t=104099)

vulture 14 November 2021 15:43

Hi Toni,

One thing I've noticed with all the v4.9.0 betas is that if a game has "Centering = vertical" then the following happens:

KS ROM (no disk inserted):
https://i.imgur.com/Ba1lgVW.png

As opposed to v4.4.0:
https://i.imgur.com/4vVHmfr.png

This affects some games, Afterburner (Sega) for example:

Afterburner (Sega):
https://i.imgur.com/r4SFwdc.png

As opposed to v4.4.0:
https://i.imgur.com/V3bDqYs.png
"G-LOC - R360 Air Battle" is also affected:

v4.9.0:
https://i.imgur.com/DMXtXZz.png
https://i.imgur.com/EL8IvGA.png

v4.4.0:
https://i.imgur.com/9hfjhOZ.png
https://i.imgur.com/RY2ctmV.png

If you need configuration files then please let me know?

vulture 14 November 2021 18:14

And another one Toni,

I was testing some games in DamienD's collection.

There's an issue with "Moon Child [AGA]" in the betas; all is fine in v4.4.0 stable.

The problem is with "MoonChildFireTest":

https://i.imgur.com/iDrCrSL.png

This whole sequence now crawls; probably 2 to 3 times slower than what it should be.

I've zipped the configuration file and <gamesdir> and put into the Zone!

Toni Wilen 16 November 2021 18:22

Quote:

Originally Posted by vulture (Post 1517031)
One thing I've noticed with all the v4.9.0 betas is that if a game has "Centering = vertical" then the following happens:

4.9 behavior is correct (at least it is technically correct!), there really is no border color when display is moved down (due to blanking) and earlier version simply made border region larger.

I am not sure if this behavior can be supported anymore due to new vertical blanking updates.

I'll check..

Zarnal 16 November 2021 19:52

Hello,

since b40 RC1, old issue ( before new chipset emulation ) with S(pi)d (64k Intro) by Zomo is back ( but slighty different ). A " ghost " scene appears just before credits ( pic 1 ). Ok with b39, b38... :)

Edit : Janaway's link is no longer working.

Edit 2 : Quickstart A1200.

vulture 16 November 2021 20:37

Thx Toni, would be nice if it could be done this way, looks better! Also, regarding the "MoonChildFireTest" issue I reported, it starts from beta 25; all is good in beta 24.

Quote:

Originally Posted by Toni Wilen (Post 1517440)
4.9 behavior is correct (at least it is technically correct!), there really is no border color when display is moved down (due to blanking) and earlier version simply made border region larger.

I am not sure if this behavior can be supported anymore due to new vertical blanking updates.

I'll check..


Toni Wilen 20 November 2021 08:59

Quote:

Originally Posted by vulture (Post 1517475)
Thx Toni, would be nice if it could be done this way

It can't work like that anymore (it is more like bug in old centering, most likely some even older version didn't work like that). Perhaps some extra option that "extends" border colors would work but it won't happen soon.

Quote:

Also, regarding the "MoonChildFireTest" issue I reported, it starts from beta 25; all is good in beta 24.
I can't notice any difference between 24 and 25. It is very slow if full cycle-exact, not that slow if only memory cycle-exact.

vulture 20 November 2021 18:29

Yup, you're right. DamienD's configuration has no "Cycle-exact" options ticked.

Ticking "Cycle-exact (DMA/Memory accesses)" gets this working like it used to in v4.4.0.

Many thanks!

Toni Wilen 21 November 2021 18:19

https://download.abime.net/winuae/fi...uae_4900b41.7z
https://download.abime.net/winuae/fi...e64_4900b41.7z

Beta 41 (RC2):

- Sprite processing was exited early if sprite horizontal start was zero (which is correct normally). But when combined with FMODE SSCAN2-bit, only "original" sprite is invisible but "doubled" sprite is still visible. Fixes Fantastic Dizzy CD32 background glitch.
- Improve Denise/Lisa internal bitplane pipeline logic, previously undocumented combined plane change + bitplane shift changes mid scanline are now working correctly.
- Above behavior also revealed yet another undocumented chipset feature/bug: if ECS Denise or AGA and bitplane plane count is lowered mid scan line and disabled plane's last bit out of bitplane shifter was one: last pixel appears twice. OCS Denise is unaffected. This is emulated only in subpixel mode.
- Clear all internal display emulation buffers at reset, previously some of them was not cleared. (Possibly could have caused temporary on screen garbage when loading statefile on the fly)

hexaae 22 November 2021 09:29

1 Attachment(s)
Quote:

Originally Posted by hexaae (Post 1507522)
Something is broken (even before b34) with display > Windowed: Filter with mask seems applied twice here in DX11, and even if Filter: Masks=none... Moreover there's a sort of strange bad buffering where I can see 2 frames constantly swapped (hard to explain)...
This happens in Windowed mode only.
https://i.ibb.co/yR6QcS1/image.png


EDIT: mmmh, seems more complicated to reproduce: only happens when you switch from Native Fullscreen to Native Windowed
EDIT2: from my early testings it started with b32

This bug is still there...

Repro steps:
- use my cfg in signature (I'm using the mask in attachment for the scanline effect...)
- run a game like Beast 1, call the UI with F12 and switch from Fullscreen to Windowed mode
the ugly effect persists even if I remove the mask (select 'none') at this point

Toni Wilen 22 November 2021 19:57

Quote:

Originally Posted by hexaae (Post 1518418)
This bug is still there...

Repro steps:
- use my cfg in signature (I'm using the mask in attachment for the scanline effect...)
- run a game like Beast 1, call the UI with F12 and switch from Fullscreen to Windowed mode
the ugly effect persists even if I remove the mask (select 'none') at this point

winuae.7z updated which should fix it, please confirm. I was confused about the mask. Does it have something to do with this bug or not?

hexaae 23 November 2021 03:22

Quote:

Originally Posted by Toni Wilen (Post 1518520)
winuae.7z updated which should fix it, please confirm. I was confused about the mask. Does it have something to do with this bug or not?

Fixed :great
Well I noticed it because I use a png as a mask for the scanline effect.
Switching Fullscreen <---> Windowed behaves as expected now without windowed artifacts.

amilo3438 24 November 2021 15:48

Animotion (Demo) by Phenomena --> http://janeway.exotica.org.uk/release.php?id=317

[A500 ECS 512 + 512] Broken from WinUAE Beta 41 (RC2)!

Toni Wilen 24 November 2021 17:40

Do you mean HAM animation part? (I even had test statefile from that part because it has been a problem long time ago)

That was one of the most stupid bug ever. Comparison that should have been "if this variable is same as that variable and another variable is zeroed" but it was "if this variable is same as that variable or another variable is non-zero"..

Which made some BPLCON0 changes not reach display rendering part (HAM/DPF/EHB on/off mainly). Fixed.

amilo3438 24 November 2021 18:33

I don’t understand too much about bed bugs, I just find them and leave everything else to the experts.. :) (confirmed to work again)

DesertBlizzard 26 November 2021 19:21

1 Attachment(s)
Anything in this dmp that might be useful to determine the reason for a random crash experienced while doing nothing other that listening to some internet radio?

Toni Wilen 27 November 2021 13:35

Quote:

Originally Posted by DesertBlizzard (Post 1519043)
Anything in this dmp that might be useful to determine the reason for a random crash experienced while doing nothing other that listening to some internet radio?

I got similar report few days ago (doing nothing and unexpected crash). It was fixed today (at least I think it was). Next (and hopefully last or second to last) beta will be released today or tomorrow.

Chipset emulation related buffer slowly fills in RTG mode (It shouldn't be even used if RTG is active and chipset display is not visible) and after about ~100 minutes it starts overwriting other variables.

hexaae 27 November 2021 15:15

This could explain also my strange random crashes since I always test it on the long run running tools, wb games, native games (mostly WHD), rtg programs... and I use an RTG Picasso96 WB.

Toni Wilen 27 November 2021 20:32

https://download.abime.net/winuae/fi...uae_4900b42.7z
https://download.abime.net/winuae/fi...e64_4900b42.7z

Official release date: about one week to go.

Beta 42 (RC3):

- Switching from D3D11 fullscreen mode to windowed mode using GUI caused windowed mode to keep-reinitializing continuously. (CTRL+F12 switch worked normally)
- Fixed wrong condition in check if Denise side BPLCON0 parameters needs updating, broke for example HAM enable/disable until next line (b41)
- Blitter line mode with C disabled also disables C-channel pointer modulo add operations. b39 update was not complete. (Methylate / Focus Design).
- Interlace mode last 2 lines (if Overscan+ or Extreme) had swapped long/short line (last line/line zero "wrap around" point), LOF toggles at this point but display device does not see it until few lines later.
- Fixed buffer overflow that caused unexpected crash due to internal variables getting corrupted. It happened if mode was RTG (native chipset not active) and native mode had copper list that writes to custom registers that can affect colors or blanking, state change was unnecessarily recorded to a buffer that is not in use if RTG mode and after long time (can be 100 minutes or more), it finally overflowed. Recent blanking updates introduced this.
- Programmed blanking change was not always fully detected, leaving partially blanked border in some situations.
- Enabled Remove interlace arfifact option without interlace mode: bitplane overrun emulation was not possible.
- Remove interlace artifact option last line flickering/garbage is really gone now. (Blanks extra line(s), this is not really the correct way but too late to fix it properly)
- Bitplane pointer pointing to upper 512k chip ram in ECS 0.5M+0.5M config didn't use indirect read routines. (Fixes Cen / Rare ECS 0.5M+0.5M config corrupted vector objects)

Viceroy 27 November 2021 21:44

You're on fire Toni. Well done!

amilo3438 28 November 2021 19:21

Quote:

Originally Posted by Toni Wilen (Post 1519166)

Beta 42 (RC3):

- Bitplane pointer pointing to upper 512k chip ram in ECS 0.5M+0.5M config didn't use indirect read routines. (Fixes Cen / Rare ECS 0.5M+0.5M config corrupted vector objects)

Not sure if this broke the "Dots (Demo) by Phenomena" in ECS... but maybe it's not important for this release. (so, just for info)

https://eab.abime.net/showpost.php?p...&postcount=852

EDIT: Quote from above link: "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."
So, maybe it is correct now because there are tall bars on top of scroller in ECS!? (before was only on the side of the screen)


All times are GMT +2. The time now is 08:40.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2022, vBulletin Solutions Inc.

Page generated in 0.12791 seconds with 11 queries