WinUAE 2.4.0 beta series

[mark_k]

While testing the latest winuae.zip I noticed a problem with Automatic resize in filter settings, which I couldn't reproduce with WinUAE 2.3.3.

The problem isn't RTG-specific. Emulated Amiga set to NTSC, ECS Agnus, OS 3.1. Workbench 640x400 NTSC high res laced, 1 bitplane.

Use Automatic resize in filter settings.
Run Pointer preferences. The Pointer prefs custom screen opens as normal (which it does when Workbench depth is insufficient to show enough colours). Drag the screen down. For me, testing under Wine 1.4rc1, The emulation window goes black with a flickering status bar, and after a short while WinUAE crashes.

Doing the same thing in WinUAE 2.3.3 there is no crash. There is however black flashing when you drag the Pointer prefs screen down from the top, but the display does reappear. And dragging it lower down the screen, there is no flashing as you drag it down.

Edit: Crash confirmed in Windows XP SP3 too.
Edit 2: When testing, I was running a utility which enables the ECS border-blanking feature.
Edit 3: I got a crash running the DOC Demons demo too. With WinUAE 2.3.3 and automatic resize enabled, it kind of "goes nuts" (window continually changes size, contents mostly black) but doesn't crash.

[Toni Wilen]

Low latency vsync is getting rewritten (again). No need to test it until next beta.

- no buffer stays the same (more or less)
- double/triple buffer should work much better, in best case there is no need for triple buffer anymore.

btw, winuae.zip is alpha, not supported except for specific test cases, only when asked.

[Mequa]

Will it be possible to use low latency Vsync with HQ filters AND fastest possible CPU?

[Toni Wilen]

Originally Posted by Mequa

Will it be possible to use low latency Vsync with HQ filters AND fastest possible CPU?

If you buy a better PC first.

[Mequa]

Core i7 @ 4GHz not up to it?

[Toni Wilen]

Short question without information = short answer without information.

ADDED:

"Yes" would be the answer to your original question but I am sure it would have not helped any more than my previous answer.
What are you trying to do and how?
Do you have problems using it? What kind of problems?
Why the smiley? (That was the reason for "buy a better PC") It didn't make any sense to me.
CONTEXT NEEDED!

HQ2X software filter has HUGE CPU usage requirements, EXTREMELY HUGE if larger than tiny resolution. Shader filter is better choice but you still need modern enough GPU.

[Mequa]

Since low latency Vsync is being re-written for double/triple buffered modes (which seem a requirement for fastest possible CPU), it seems more fruitful to test this when the next beta arrives. I was just asking if this combination would be officially supported on high-end CPUs in next stable release. (Of course HQ2X will choke an Intel Atom etc.)

[Toni Wilen]

Originally Posted by Mequa

Since low latency Vsync is being re-written for double/triple buffered modes (which seem a requirement for fastest possible CPU), it seems more fruitful to test this when the next beta arrives. I was just asking if this combination would be officially supported on high-end CPUs in next stable release. (Of course HQ2X will choke an Intel Atom etc.)

Ok, finally your questions makes some sense
There is no "official" support because it depends heavily on PC's performance. (and I still hate most filters. Also software versions are quite stupid today)

Also there has never been any "official" support for "Does it perform without slowdowns?" Or something like that. Only "official" answer is: Buy a Better PC!

[Toni Wilen]

http://www.winuae.net/files/b/winuae_2400b16.zip

Beta 16:

- AROS CDFS update, disabled annoying and pointless "Write protected" requester.
- Reload shader filter automatically if shader file's modification date changes, quicker to experiment with filters, changes will be visible immediately.
- Added texelsize as a parameter to _winuae.fx shader, needed to quickly access other pixels (not just current pixel), version number bumped to 3.
- DirectDraw mode screen blanked when adjusting slider settings on the fly.
- Reordered emulated autoconfig boards, first comes all Z2 boards and then all Z3 boards. KS1.x stops autoconfig process when first Z3 board is found, previously filesystem "board" was after Z3 boards causing UAE controller HDs to not mount under 1.3 if Z3 or RTG was enabled. Debugger 'dm' command also shows name of current autoconfig board in autoconfig config space if autoconfig didn't finish.
- Picasso_FindCard() will now return error if RTG board is not fully autoconfigured.
- Added RTG Z2/Z3 board mode selection. NOTE: both Z2 Fast RAM and Z2 RTG RAM must fit in 8MB Z2 autoconfig region.
- Removed NO_SHUTUP flags from emulated autoconfig devices, AOS is now allowed to "shut up" the card (usually can only happen if there is no free config space) and continue autoconfig process. Probably never happens but maybe someday it is needed.
- Clear RTG screen buffer when switching modes, sometimes previous graphics was visible on border area if RTG resolution changes without host screen mode change.
- Low latency vsync partial rewrite, fastest possible CPU/JIT should have better peformance compared to previous betas, triple buffering is not needed anymore, double and triple should have exact same performance, triple only increases latency. RTG low latency vsync should be stable now, some hangs also fixed.

NOTE: Previous betas saved gfx_backbuffers_rtg=2 to configuration file, edit it or remove it manually if you want to use double buffer (=1) RTG mode which is also the default now. There is no GUI option yet.

Always include both logs and config file if you have any kind of vsync related problems, thanks!

[arti]

Hi Toni,

when option "Pause emulation" is checked in minimize or inactive window an alt key stays pressed.

Using AmiKit 1.6.1.

Best regards

[FreakyDan]

Latest beta16 introduced screen tearing again for me using low latency with no buffer. Beta15 runs silky and smooth

Forgot to include the config, here it is....

[Toni Wilen]

Originally Posted by arti

when option "Pause emulation" is checked in minimize or inactive window an alt key stays pressed.

Confirmed, fixed in next beta.

Originally Posted by FreakyDan

Latest beta16 introduced screen tearing again for me using low latency with no buffer. Beta15 runs silky and smooth

Only in interlaced mode? (Log shows there is something wrong with interlace)

EDIT: Tearing problem found but what did you do to cause all those mode change log lines? Some kind of special screenmode?

[arti]

Also, it would be better to auto unpause when window is again active.

[FreakyDan]

Originally Posted by Toni Wilen

Only in interlaced mode? (Log shows there is something wrong with interlace)

EDIT: Tearing problem found but what did you do to cause all those mode change log lines? Some kind of special screenmode?

It's the demo Storm by Impulse. Don't know why the demo behaves like that when showing an interlaced picture. A couple of versions back of winuae the problem could be fixed by using kick3.0 instead of kick3.1. But with the latest version it doesn't matter which kick I use.

[Dr.Venom]

Originally Posted by FreakyDan

It's the demo Storm by Impulse. Don't know why the demo behaves like that when showing an interlaced picture. A couple of versions back of winuae the problem could be fixed by using kick3.0 instead of kick3.1. But with the latest version it doesn't matter which kick I use.

It's giving "40-815 [2330 000x000]: Interlaced frame type mismatch 1<>0" continuously. Happens in both fastest possible and cycle exact..

Edit:

This seems to be a demo issue. The above only happens when using Kickstart 3.1. The interlaced screen appears fine for me when using Kickstart 3.0.

I remember this to be an issue with more demos back in the days , where interlace would work fine on 3.0, but would show long and short frame on top of each other (or alternating the fields in the wrong order?) when using 3.1. Luckily I had a Blizzard 1260 with the kickrom functionality, so I could always reverse to 3.0. But still it was a bit of a pain in the ass..

@ FreakyDan, I tested this with the latest unofficial Beta http://www.winuae.net/files/b/winuae.zip (exe dated 11-feb 17:47) . Could you try that one?

[Toni Wilen]

Originally Posted by Dr.Venom

This seems to be a demo issue. The above only happens when using Kickstart 3.1. The interlaced screen appears fine for me when using Kickstart 3.0.

Yes, it is. Demo takes over the display without taking over the system properly. 3.1 ROM code keeps setting VPOSW LOF bit each frame while demo uses interlace mode. This combination isn't going to work very well.. Not sure why 3.1 introduced this extra code.

[Toni Wilen]

Originally Posted by arti

Also, it would be better to auto unpause when window is again active.

It does except magic mouse seems to not unpause it for some reason. Will be fixed..

[Dr.Venom]

Originally Posted by Toni Wilen

Yes, it is. Demo takes over the display without taking over the system properly. 3.1 ROM code keeps setting VPOSW LOF bit each frame while demo uses interlace mode. This combination isn't going to work very well.. Not sure why 3.1 introduced this extra code.

Good to have this confirmed and also know the reason why this is happening. The question remains how to deal with it in emulation. In my recollection these type of demos would still show a (wrongly interlaced) picture on 3.1, while WinUAE now sort of "hangs" on the situation, which might be confusing to users.

Would it be a solution that if you detect more than 2 (or three to be safe) frame type mismatches in a row && kick 3.1 is set, that you disable the code that "skips" a frame to match frame type? That way it would not lock WinUAE, display interlace wrongly in this specific situation (just like it would on a real Amiga) but still have correct interlace display in all other situations.

[mingle]

Hi Guys,

Just giving 2400B16 a run and noticed that it seemed quite sluggish when running my Workbench (ClassicWB), so I ran a few tests using SysSpeed. See the attached screen-grab for the results.

The WinUAE config is identical for all tests, the only difference is the WinUAE executable used.

The memory and drive functions look (overall) slightly better for the 2400B15 & 16, but why is B16 showing a "0.00" value for "Cache Read"?

All intuition-based tests seemed to crawl on 2400B16.

Any ideas why?

Cheers,

Mike.

[Toni Wilen]

Originally Posted by mingle

Hi Guys,

Just giving 2400B16 a run and noticed that it seemed quite sluggish when running my Workbench (ClassicWB), so I ran a few tests using SysSpeed. See the attached screen-grab for the results.

The WinUAE config is identical for all tests, the only difference is the WinUAE executable used.

The memory and drive functions look (overall) slightly better for the 2400B15 & 16, but why is B16 showing a "0.00" value for "Cache Read"?

All intuition-based tests seemed to crawl on 2400B16.

Any ideas why?

Cheers,

Mike.

Error: "Missing configuration file error" error. (and test above winuae.zip first)