WinUAE 2.4.0 beta series

[Toni Wilen]

2.4.0 beta series.

Display handling rewrite, Fastest possible CPU/JIT low latency vsync support, A2024 and Graffiti emulation. And bug fixes.

Quick rules:

Allowed posts in this thread:

- bug reports but make sure bug is NOT in previous offical version. (Old bug -> separate thread)
- comments and suggestion about new features in betas or changed feature in beta. (missing feature or old feature which has not been changed in latest beta series -> separate thread)

Everything else will be deleted, no questions asked. (includes offtopic posts, discussion about pros and cons of public beta testing etc..)

Misc notes:

- "not working" is not a bug report -> instadelete
- always include name of the program. "Everything" or "most games" fail is not a bug report!
- always use quickstart settings. Non-QS setting reports will be ignored unless there is reason why quickstart isn't good idea but still test using quickstart just to confirm! (unless it needs some HD highend config)
- always check logs (both winuaebootlog.txt and winuaelog.txt, don't forget to enable logging in misc-panel) for possible error or warning messages.
- remember to include PC specs + Windows version (if asked)
- graphics error -> include screenshot.
- disk based program -> include CRC32 (mouse over disk history arrow down button will show CRC32) of the image(s) (or SPS ID or full tosec name if from latest set). Plain name of the game is NOT generally enough! (usually there are different cracks, trainers etc..)

As usual, betas may destroy your PC and your sanity

This thread is for general discussion (bad/good/stupid idea/whatever, no trolling or flaming) about public beta testing.

Beta test related PMs or emails will be generally ignored.

[Toni Wilen]

Finally..

http://www.winuae.net/files/b/winuae_2400b1.zip

Beta 1:

- Number of supported serial ports, MIDI ports and sound devices increased, allocated dynamically.
- Fixed crash if more than max supported serial ports or MIDI ports found.
- Sort MIDI and serial ports alphabetically.
- Hardfile write protection has never worked, only worked when using real harddrives.
- Very old keyboard handshake emulation hack removed, now following key code won't arrive until current key code has been handshaked properly. I guess this hack was needed when CIA emulation was not good enough.
- Parallel joystick adapter unused pin added to Input panel (spare/2nd button) World Cup 1990 parallel port joystick adapter uses this pin.
- Gayle IDE emulation IDE doubler mode improved, both buses are now separate, supports drivers that send command to drive on another bus when drive on other bus is still processing previous command (Linux/NetBSD/AROS)
- Emulate disk write with enabled wordsync. Write does not start until wordsync marker has been found on disk. Some games accidentally enable wordsync bit when writing. Previously write never started.
- Saving configuration crashed if scale mode was set to integer.
- -portable disables winuaebootlog.txt, also added -bootlog and -nobootlog parameters.
- Switching from physical CD to image or vice versa didn't send uaescsi.device change notifications
- Support added for special screen modes that can have different "input" and "output" resolutions or positioning. On the fly switching also supported.
- A2024 and Graffiti emulation implemented, uses above system. Configuration in chipset panel, autodetect = attempts to detect if mode is normal, A2024 or Graffiti.
- All programmed display modes (DBLPAL/NTSC, MULTISCAN etc..) should be always centered, exact display positioning configuration is now taken from sync position custom registers.
- Normal modes may also have better centering without extra configuration.
- Mode switching can leave old graphics garbage, this will be fixed later.

A2024 and Graffiti emulation notes:
- 32-bit display mode required.
- Direct3D or DirectDraw + null filter mode required. (=Does not work if DirectDraw + no filter)
- Make sure display resolution (lores/hires/shres) is equal or higher than required resolution:
* A2024 10Hz: lores
* A2024 15Hz: hires (loss of every other pixel if lores)
* A2024 only tested using WB drivers (which don't use all A2024 special features)
* Graffiti lores: hires (does not work at all if lores)
* Graffiti hires (AGA required): shres (does not work at all if hires or lores)
- A2024 output resolutions: 1024x1024 (PAL), 1024x800 (NTSC), OCS chipset: 1008x1024 or 1008x800.
- Yes, weird A2024 refresh behavior is normal, single display is built from multiple frames!
- A2024 enabled but special mode line not detected: fall back to normal native display. Real A2024 would show 8 grayscale screen. This is not (yet? Pointless feature?) emulated.

NOTE.NOTE.NOTE.NOTE: Because display positioning has been changed:
- Remove all custom filter positioning settings first!
- Also test without enabled horizontal or vertical centering.

NOTE2:NOTE2:NOTE2: This can also change again during beta period. It won't be compatible with old versions.

[Steve]

I only tested it for a short time on my machine but initial findings were that cycle-exact mode has slowed to a crawl on my machine. I am using a basic A1200 setting which works fine with the current release. The game I tested using 50Hz/No buffering was Banshee. A500/A1200 cycle exact setting makes the game run too fast and adding cycle-exact option slows the emulation to a pitiful 33 fps. The same setup gives smooth 50 fps with the current release.

[Toni Wilen]

Originally Posted by Steve

I only tested it for a short time on my machine but initial findings were that cycle-exact mode has slowed to a crawl on my machine. I am using a basic A1200 setting which works fine with the current release. The game I tested using 50Hz/No buffering was Banshee. A500/A1200 cycle exact setting makes the game run too fast and adding cycle-exact option slows the emulation to a pitiful 33 fps. The same setup gives smooth 50 fps with the current release.

Logs missing.

[Ian]

Scaling/centring now works in all screen modes (eg dblpal).

[Toni Wilen]

http://www.winuae.net/files/b/winuae_2400b2.zip

Beta 2:

- Adjusted horizontal positioning, full overscan was not visible.
- Clear display buffers when monitor type or refresh rate changes.
- A2024 1-plane mode brightness level fixed (DPL bits implemented)
- Fix A2024 10Hz flickering near panel borders.
- 16-bit color supported in A2024/Graffiti emulation.
- Disable audio emulation hacks if 68020 CE (previously only if 68000 CE)
- Switching from CE to non-CE on the fly froze the emulated machine in some situations until mode was switched back to CE.
- PCMCIA CF IDE emulation added ("SC IDE" controller type), originally developed for AROS m68k PCMCIA CF IDE support testing.
- Fastest possible without JIT + low latency vsync should work much better now (but probably still not good enough)

[Retroplay]

2.4.0 beta 1 and 2 crashes every time I try to boot my ClassicWB setup.
It's working 100% with 2.3.3.

[Toni Wilen]

Originally Posted by Retroplay

2.4.0 beta 1 and 2 crashes every time I try to boot my ClassicWB setup.
It's working 100% with 2.3.3.

As a workaround enable at least null filter or use D3D mode. Will be fixed later..

[Retroplay]

Thanks, directdraw and null filter did the trick.

[slightly beta off-topic]
If I use D3D and no filters I get slighly garbled text/graphics, this is not beta related though as it happens in older versions as well.



So that's why I still use directdraw.

[FreakyDan]

Originally Posted by Retroplay

Thanks, directdraw and null filter did the trick.

[slightly beta off-topic]
If I use D3D and no filters I get slighly garbled text/graphics, this is not beta related though as it happens in older versions as well.



So that's why I still use directdraw.

Set the scaling to "No scaling" should fix your garbled text/graphics..

[Retroplay]

Originally Posted by FreakyDan

Set the scaling to "No scaling" should fix your garbled text/graphics..

Doesn't make any difference.

Sorry for the off-topic posts in this thread, wont happen again.

[edit]
Integer scaling did the trick.

[Toni Wilen]

Originally Posted by Toni Wilen

As a workaround enable at least null filter or use D3D mode. Will be fixed later..

Does http://www.winuae.net/files/b/winuae.zip fix the crash? (It should but there can be some conditions I missed..)

[Retroplay]

Yes that fixed it.
Many thanks.

[lordofchaos]

Just ran beta 2400b2 and have encountered a bug/issue whilst running Uridium 2 "WHD_UridiumII release 0801.1" using WHDLoad version 16.9

The main title screen module seems to be corrupted with raw data making it sound pretty ugly, I have also run the game with the same config on WinUAE 2.3.3 and the raw data is absent and the sound appears to be working correctly.

Also I have run the game on my real A1200 (68030, 32 Meg fast) using the identical WHDLoad version and title tune plays correctly without corruption.

Log, Config and save state included.

*Update 1

Some more strange behavior, I just fired up WinUAE (2400b2) again and used my A1200 Config (attached) but decided to change Z3 Fast from 32 Meg to 16 Meg and then ran Uridium 2 WHDload, title screen music played fine this time. This seemed strange so I quit WinUAE, reloaded and repeated the sequence but this time the game title track was corrupt again. I did this sequence at least 6 times and the results would vary. Random memory issue?

Going to try and replicate this issue on the my real A1200 by running game multiple times, could be a issue that occurs on real hardware and with WHDLoad install it`s self?

*Update 2

I can not replicate this issue on real hardware (A1200 68030 32MB FAST).

*Update 3

Just did some more testing using Beta 2, 2011.12.08 (One posted in moonstone sound issue thread).

Starting WinUAE with 1200 config but unticking CE yeilds the following results: Games main title score & Chip loader score will play properly without sound corruption, I restarted and tried this sequence 3 times in a row.

Starting WinUAE with 1200 config with CE ticked yields the following: Chip sound loader module (the copper bands one) plays with sound grating, main title module still has garbaged sounds.

[Ian]

Massive slowdowns when using JIT and/or fastest possible mode, introduced in Beta 1, still present in beta 2.

Installing Turrican WHD is a good example to use I guess.

I know JIT isn't for games, but it does cause slowdowns in all games I tried.

Attached is logs and config I normally use for 3.9. Obviously I removed cycle exact and set fastest possible and JIT for the purpose of the test.

[Toni Wilen]

Originally Posted by Ian

Massive slowdowns when using JIT and/or fastest possible mode, introduced in Beta 1, still present in beta 2.

It is normal, low latency vsync + JIT has never been supported. (check changelogs). It was accidental if it worked for you, normally it halves the speed or something like that.

[Toni Wilen]

Beta 3:

http://www.winuae.net/files/b/winuae_2400b3.zip

Long waited fastest possible CPU vsync is here! Maybe.

- Fixed DirectDraw mode crash (b1)
- Fixed uaenet.device crash if program attempted to open it but winpcap was not installed. (old)
- Added CD32 drive emulation hack, CD32 Mutation Gold Compilation games now load (but they didn't work on my real CD32 either so technically not working is correct behavior. This hack is removed if it breaks other programs)
- Too fast CPU audio hack update, one of the hacks actually caused sound glitches with some strange sound routines.
- Do not blank screen if display can't be shown (DirectDraw display position negative upper left coordinates), now sets coordinates to zero.
- Keyboard statesave buffer overflow (b1)

- Experimental new vsync mode for fastest possible/JIT CPU modes. Tested using WHDLoad demos and games.
* - Uses secondary thread which polls vblank state continuously. There is no way to get signal/message/whatever during vblank. Stupid Windows.
* - Activates automatically if 2 or more CPUs/cores detected and fastest possible/JIT and low latency mode selected.
* - JIT is also supported.
* - Works in windowed and fullscreen modes (just like normal low latency vsync).
* - This mode is currently buffered, used to hide horrible tearing (flip timing is much more complex if fastest possible CPU), will be hopefully improved later.
* - Sound pitch changes possible, this also needs some improving later.

[lordofchaos]

Regarding previous Uridium 2 sound issue post:

Have just tested the game again with A1200 config but raw data is still present with most recent Beta 3. However changing the cpu from 68020 to 68030 appears to fix the problem. Have not tested this with previous beta`s yet.

[Toni Wilen]

Originally Posted by lordofchaos

Regarding previous Uridium 2 sound issue post:

Have just tested the game again with A1200 config but raw data is still present with most recent Beta 3. However changing the cpu from 68020 to 68030 appears to fix the problem. Have not tested this with previous beta`s yet.

I can't duplicate. I tried both 68020 fastest possible and cycle exact. Title tune worked fine.

But perhaps it is version specific, there are 3 different 0.801.1 WHD UridiumII versions.

[lordofchaos]

But perhaps it is version specific, there are 3 different 0.801.1 WHD UridiumII versions.

I have attached the version on my HD. It does state it`s 0.801.1.

[Toni Wilen]

Thanks. Answered in moonstone thread.

[Dr.Venom]

Originally Posted by Toni Wilen

Beta 3:

http://www.winuae.net/files/b/winuae_2400b3.zip

Long waited fastest possible CPU vsync is here! Maybe.

- Experimental new vsync mode for fastest possible/JIT CPU modes. Tested using WHDLoad demos and games.

Awesome . First impressions are that it is working very well. Workbench stuff is working really fast, and most whdload games and demos run without problems. Specific cool things: Just ran the TBL - Starstruck demo and it's running really very very smoothly... . End scroller part of Big Time Sensuality by Axis is also flawless.

* - This mode is currently buffered, used to hide horrible tearing (flip timing is much more complex if fastest possible CPU), will be hopefully improved later.

Hopefully that's possible, as the latency introduced is very noticable for faster games (like eg Pinball Dreams..) when compared to the unbuffered cycle exact mode.

* - Sound pitch changes possible, this also needs some improving later.

Only noticable sound issue I found up till now is the WHDLoad version of Jim Power, where the high pitched sound of your bullet hitting enemies is off-pitch in fastest possible mode. Running the same config with only changing 'fastest possible' mode to 'cycle exact' mode the sound is OK.

[Steve]

I tried WHDLoad Breathless and it completely crawled. Tried windowed mode and it worked better until I made the window larger to see the display, then it crawled.

Quick question that probably isn't beta related. What has happened to keyboard input?? Is it broken? Why do I have to select 'Game Ports Panel' in the Input section to get the keyboard working again, thus losing my custom configured game pad settings? What has happened? Is it like this by design now??

Tried with standard non-JIT low-latency fs config and received mixed results with this beta compared to the last version. CPU usage is now 90-100% compared with 40-50% in the previous official release. Has my CPU fan whirling round like a banshee. Setting JIT makes no real difference although if I also switch on the fastest possible mode then it really slows down and CPU usage is upto 150%. Is this because it is just a debug executable or maybe somthing else?

[msayed1977]

- Mode switching can leave old graphics garbage, this will be fixed later.

Please don't forget to fix very soon.

[Toni Wilen]

Quick question that probably isn't beta related.

Do a quick test first before deciding if it is or not beta related..

Tried with standard non-JIT low-latency fs config and received mixed results with this beta compared to the last version. CPU usage is now 90-100% compared with 40-50% in the previous official release. Has my CPU fan whirling round like a banshee. Setting JIT makes no real difference although if I also switch on the fastest possible mode then it really slows down and CPU usage is upto 150%. Is this because it is just a debug executable or maybe somthing else?

Non-fastest CPU modes should be the same, perhaps something got broken.

Fastest possible or JIT vsync requires much more CPU power. Perhaps even a triple/quad core CPU so that both main UAE thread and vblank thread get near 100% CPU time. Unfortunately there really isn't any other way to make JIT+vsync to work perfectly. Fastest possible without JIT can work (it did in b2, this beta uses new mode for both. Future versions may have an option).

I tested this on my laptop and it didn't seem to cause any more fan sound than normal fastest possible CPU mode. Buy a better cooler or PC

Whats your hardware specs?

Please don't forget to fix very soon.

It should be fixed. You sure you aren't confusing this with buggy programs with missing blitter waits? These commonly cause random crap if emulation configuration isn't exact enough.

[Steve]

No it wasn't beta related. Stilll weird though.

Dual Core Pentium 2.13Ghz. Geforce 9800GTX+, 2GB RAM, 750watt PSU. No extra cooling systems!

[Dr.Venom]

Originally Posted by Steve

I tried WHDLoad Breathless and it completely crawled. Tried windowed mode and it worked better until I made the window larger to see the display, then it crawled.

FWIW (this isn't to say there isn't an issue with the beta), but I can't confirm this slowdown on my setup. Breathless runs mostly smoothly in full-screen mode up to the largest screen/window setting within Breathless config.

Tried with standard non-JIT low-latency fs config and received mixed results with this beta compared to the last version. CPU usage is now 90-100% compared with 40-50% in the previous official release.

Can't confirm this either. Both B2 and B3 consume about 40% CPU (as shown by the onscreen LED) on my setup when using non-fastest CPU modes, much the same as the previous official release. When using fastest possible, both B2 and B3 show onscreen leds at 100% cpu usage. Are you sure you didn't accidentally compare a fastest possible with a non-fastest possible config?

Has my CPU fan whirling round like a banshee. Setting JIT makes no real difference although if I also switch on the fastest possible mode then it really slows down and CPU usage is upto 150%. Is this because it is just a debug executable or maybe somthing else?

When comparing non-fastest (cycle-exact) to fastest mode my CPU fan will also make more noise, but to me that's normal since the CPU is stretched more. Also in my setup there's no difference in additional fan whirling between the fastest possible modes (non-vsynced) in =<2.3.3 releases and the vsynced fastest possible modes in these beta's. They both whirl up the fan a bit. This is on a Core I7-920 overclocked to 4Ghz with a Scythe Mugen 3 cooler (which when not stressed runs at about 400 RPM )

Toni, I'm not sure, what is the expected behaviour for the "CPU Idle" setting/slider in these beta's? Should it be able to lower the 100% cpu usage in the onscreen led to something less when used? Currently the setting doesn't seem to have impact, but maybe I'm missing something (never made much use of this setting..)

[rsn8887]

When using non-latency Vsync without cycle exact and any speed setting "Fastest Possible/approximate A1200/adjustable CPU," I am getting jerky scrolling and pretty terrible screen tearing. Tested with Beta 3 and Hyperdrive (PD-Game). Problem remains when setting priority to above normal, increasing cpu idle or switching buffering between no/double/triple.

Logs are attached.

Scrolling is perfect in non-latency Vsync mode when using cycle exact. Sadly, some other games have slow down in this mode, I think because my PC is not fast enough for cycle exact A1200 mode when there's lots going on (e.g. in Slamtilt).

I have a dual core machine. It is a 2.8 GHz Pentium D. I am using Directdraw and 100 Hz (PAL). I tried briefly using Direct3D, but got a blue screen.

Dump from that Direct3D crash is also attached.

This might be related: Before non-latency Vsync was implemented, the old Vsync actually worked for me in fastest possible. It was weird though, because CPU usage was NOT at 100%. Since non-latency Vsync was implemented, the old Vsync has jerky scrolling now.

EDIT: When I use Directsound instead of Portaudio/ASIO4ALL, the scrolling is not jerky, there's some sound pitch shifting, and there's still tearing. I suppose the new vsync mode is related to sound? If I disable sound emulation, the emulator runs at 400 fps!

[Toni Wilen]

Originally Posted by Dr.Venom

Toni, I'm not sure, what is the expected behaviour for the "CPU Idle" setting/slider in these beta's? Should it be able to lower the 100% cpu usage in the onscreen led to something less when used? Currently the setting doesn't seem to have impact, but maybe I'm missing something (never made much use of this setting..)

I am not yet sure if it even can work in vsync modes

Originally Posted by rsn8887

Scrolling is perfect in non-latency Vsync mode when using cycle exact. Sadly, some other games have slow down in this mode, I think because my PC is not fast enough for cycle exact A1200 mode when there's lots going on (e.g. in Slamtilt).

Normal vsync modes can be broken, later..

I tried briefly using Direct3D, but got a blue screen.

Update your display drivers. DirectDraw support isn't update yet. It is always D3D that comes first, when it is considered working, DirectDraw gets updated. (I don't want to rewrite both code paths multiple times)

EDIT: When I use Directsound instead of Portaudio/ASIO4ALL, the scrolling is not jerky, there's some sound pitch shifting, and there's still tearing. I suppose the new vsync mode is related to sound? If I disable sound emulation, the emulator runs at 400 fps!

Only DirectSound has autoadjustment (+pitch changes). Sound tweaks will come later.

It is always good idea to use normal defaults when testing new features (DirectSound, D3D) and then compare differences.

[Toni Wilen]

http://www.winuae.net/files/b/winuae_2400b4.zip

Beta 4:

- Another switching from CE mode to non-CE hang fix. (Blitter nasty related)
- Another too fast CPU audio hack update, fixes random glitches (Uridium II, Moonstone, probably many others)
- Low latency vsync buffering change, selected buffering mode can be modified now.
* - No buffering is same as previously
* - Double and triple buffering can be enabled, increases latency but can reduce random jumps.
- Fastest possible CPU low latency vsync update:
* - Works now without buffering
* - Flips buffers in vblank thread (Flipping timing is correct now, tearing gone)

[rsn8887]

Originally Posted by Toni Wilen

It is always good idea to use normal defaults when testing new features (DirectSound, D3D) and then compare differences.

Ah, thanks for clearing that up.

I just tested newest beta 4 using Directsound and Direct3d.

Tearing is gone, scrolling is mostly smooth.

However, there are still random stutters in the scrolling, jerky scrolling, usually one or two per second, sometimes much more frequent, sometimes less. It seems to depend on how much is going on in the game, but I am not sure about that. I tested mostly with Slamtilt, but also with Hyperdrive. Much less is going on in Hyperdrive, still see some stutters. Stuttering is worse at low sound buffer settings (1,2), but does not go away at larger buffer settings. Double or triple buffer doesn't change it.

I am attaching the logs.

[Toni Wilen]

Another test: http://www.winuae.net/files/b/winuae.zip

This version does not lose frames even if vblank detection gets delayed for a few milliseconds (one PAL frame is only 20ms, NTSC is ~16ms). Delayed vblank now causes tearing at the top of screen which is less annoying than jumps.

Try following base configuration, when/if it works, then try your custom configuration(s):

- low latency vsync
- directsound, buffer 4
- 50hz or 60hz refresh rate (100Hz+ may not work yet) Test also 60Hz because sometimes 50Hz simply won't work even if monitor supports it.
- fullscreen (windowed usually also works but Aero can cause issues in some situations)
- try all buffering setups
- cycle-exact (a500 or 1200), fastest possible and JIT. (Interestingly JIT may work better than non-JIT fastest possible!)

[rsn8887]

I only tested 100 Hz so far (cannot test 50 Hz) and a bit of 60 Hz. Fastest Possible non-JIT.
no buffering: works fine for a while (10 secs), then stutters
double buffering: works fine for while, then screen tearing
triple: seems perfectly fine so far. Stays fine with directsound, sound buffer=2. Tried Portaudio Async4All, but starts stuttering as soon as Portaudio is enabled, regardless of sound buffer setting.

In 60 Hz, I got erratic "flickering" in all buffering modes (rate about 2 Hz). With flickering I mean that for example in slamtilt, an old frame flickers together with the new frame. If only the ball is moving, it looks like it has a ghost ball behind it that flashes every once in a while.

PS: once it starts stuttering or tearing or flickering it doesn't really stop until I go to options, change d3d filter or priority or sth, and click ok, then it is usually good for a little bit. It is rather erratic though.

PPS: In JIT mode the game slamtilt crashed on the intro screen.

[Toni Wilen]

Does it work smoothly if you disable sound completely? (Disable it before starting emulation). What does SND% show? Does it flicker yellow or blue?

Can anyone else (than me and Dr Venon) make this work? It works more than perfectly here, soooooo smooooth whatever the mode (CE/fastest possible/JIT), even in windowed mode.

[rsn8887]

Originally Posted by Toni Wilen

Does it work smoothly if you disable sound completely? (Disable it before starting emulation). What does SND% show? Does it flicker yellow or blue?

Can anyone else (than me and Dr Venon) make this work? It works more than perfectly here, soooooo smooooth whatever the mode (CE/fastest possible/JIT), even in windowed mode.

Yes it works perfectly in all buffering modes, if I disable sound completely. With sound on, it only works stutter free in triple buffered mode. The snd% is around +12 (or -12) hard to tell, it flickers, in all modes. It is not blue or yellow.

(ignore this it is untrue:
It seems to be related to how much is going on in the game. I tested hyperdrive instead of slamtilt, and it looks like it works smooth without stutters even in "no buffering" mode.
)

Damn, I was still in triple buffering when I tested with hyperdrive. It does stutter when sound is on, unless I go to triple buffering after all.

[Toni Wilen]

Originally Posted by rsn8887

Yes it works perfectly in all buffering modes, if I disable sound completely. With sound on, it only works stutter free in triple buffered mode. The snd% is around +12 (or -12) hard to tell, it flickers, in all modes. It is not blue or yellow.

Strange. No colored flashing = no buffer under or overflow but it still seems to cause similar problems.. Could you attach your config file? Perhaps you have some setting enabled that is not handled correctly.

Any change if you try different active priority settings?

[rsn8887]

Originally Posted by Toni Wilen

Strange. No colored flashing = no buffer under or overflow but it still seems to cause similar problems.. Could you attach your config file? Perhaps you have some setting enabled that is not handled correctly.

Any change if you try different active priority settings?

There's no change when I try different priorities. I am attaching the config.

I think it is weird that it really works perfectly fine in triple buffered mode (as long as I use Directsound and not portaudio), just not in no buffered and double buffered.

[Toni Wilen]

Originally Posted by rsn8887

There's no change when I try different priorities. I am attaching the config.

I think it is weird that it really works perfectly fine in triple buffered mode (as long as I use Directsound and not portaudio), just not in no buffered and double buffered.

Triple buffering is the only buffering mode that can work without slowdowns if page flipping is not immediate, perhaps you have forced vblank wait set in display driver control panel utility?

[rsn8887]

Originally Posted by Toni Wilen

Triple buffering is the only buffering mode that can work without slowdowns if page flipping is not immediate, perhaps you have forced vblank wait set in display driver control panel utility?

My nvidia control panel (I have a Geforce 9500 GT) is at default settings, using latest driver version. The default for "Vertical Sync" is "Use the 3D application Setting," I just tried to force it to off, but same behaviour in WinUAE.

[Toni Wilen]

I tried WHDLoad SlamtiltAGA, perfect smooth scrolling. (with or without JIT)

Could you try Vision Megademo IV? It is my favorite vsync test program because it has lots of parts with big scrollers (For some reason I need to have JIT enabled when starting it or it crashes.. after it has started JIT can be disabled)

winuae.zip also updated again, some sound adjusment tweaks (must use DirectSound)