WinUAE 5.1.0 beta series

[Ponki1986]

Originally Posted by Toni Wilen

- Built-in (statically linked) Munt v2.7.0 MT-32/CM-32L emulator. Listed in MIDI out menu. ROM images needed, either in c:\mt32-rom-data\ or in <winuae rom path>\mt32-roms\. Uses same ROM file names and ROM version selection priority order as DOSBox. Compressed ROMs are not supported. MIDI audio is mixed with Paula audio.

It works fine for me under Windows 7 x86. But why has the Microsoft GS Wavetable Synth disappeared? Additionally, efficiency decreased. Version 5.0 achieved approximately 430 MIPS on SysInfo. Now I'm about 190. This is also visible in AmiQuake where TimeDemo gives a result of 16 FPS. Under version 5.0 I had 23 FPS.

[Toni Wilen]

Originally Posted by Ponki1986

It works fine for me under Windows 7 x86. But why has the Microsoft GS Wavetable Synth disappeared? Additionally, efficiency decreased. Version 5.0 achieved approximately 430 MIPS on SysInfo. Now I'm about 190. This is also visible in AmiQuake where TimeDemo gives a result of 16 FPS. Under version 5.0 I had 23 FPS.

Last MIDI Out device was missing, fixed in next beta.

About speed: more details needed (I don't usually care too much about speed unless someone includes detailed info and/or it is very noticeably difference. Compatibility always comes. The more complex code that more slowdown, especially if CPU is relatively old with small caches. "Buy a new PC" and so on But seriously, this speed difference is too big, there must be some real reason.).

Does it happen:
- only when native display mode is visible? Not when RTG mode? This is very important bit. (Native mode got more complex to emulate again which can explain part of the problem but probably not fully)
- same relative difference with and without JIT?
- vsync and every filter mode must be disabled.
- audio emulation disabled, any difference?
- nothing weird in log?

Originally Posted by 8bitbubsy

Ah I see, didn't know it was that sensitive to small code changes. I'll stop reporting it in the future.
I guess I'll start using a P96 video mode instead for my development environment.

This time it was quite simple fix but it will break in the future again. If you only need native mode with higher vertical resolution, try DBLPAL/NTSC modes.

[Toni Wilen]

https://download.abime.net/winuae/fi...nuae_5100b2.7z
https://download.abime.net/winuae/fi...ae64_5100b2.7z

Beta 2:

- If configured monitor size/coordinates changed, ignore it if it is not connected to any RTG board. Extra windows opened if non-existing monitor's window size was modified on the fly.
- Some paths were shown strangely in floppy/quickstart floppy paths.
- Bitplane start was horizontally shifted in some situations.
- Support autovectored interrupts in MMU and generic fast CPU modes (Hatari)
- Fixed Munt (and also other sound streams that mix with Paula audio) stopping if sound settings were changed.
- Last enumerated MIDI out device was missing.
- Fixed "Remove interlace artifacts". (Missing reset of separated Denise and Agnus horizontal counters)

Help to narrow down reported slow down is accepted (for example if it only happens in some specific configs, some option enabled or not enabled etc)

[Ponki1986]

Originally Posted by Toni Wilen

- Last enumerated MIDI out device was missing.

The Microsoft GS Wavetable Synth is back, thank you very much.

However, in the case of efficiency, I would like to apologize. I raised the alarm too quickly because after a Windows reset, performance is good. Yes, there are a few fewer MIPS, but I understand it because new functions are being introduced.

[Ponki1986]

I think I found another little bug. When I run a program on a separate screen, e.g. Personal Paint, the mouse cursor disappears after finishing its work and returning to Workbench. A similar thing happens when you run the Matrix 1.9 screensaver from Olivier Fabre. I have to run Pointer Prefs and click use setting to get the pointer back. I did not notice this problem under WinUAE version 5.0 and lower.

Additionally, I noticed that when listening to CDDA music using PlayCD, from the mounted BIN/CUE image, there is silence for a second when I switch screens. I have not encountered such a situation in previous versions.

All of the above applies to Beta 2. I haven't checked Beta 1.

[Toni Wilen]

Originally Posted by Ponki1986

The Microsoft GS Wavetable Synth is back, thank you very much.

However, in the case of efficiency, I would like to apologize. I raised the alarm too quickly because after a Windows reset, performance is good. Yes, there are a few fewer MIPS, but I understand it because new functions are being introduced.

Nice. And weird side-effect..

Originally Posted by Ponki1986

I think I found another little bug. When I run a program on a separate screen, e.g. Personal Paint, the mouse cursor disappears after finishing its work and returning to Workbench. A similar thing happens when you run the Matrix 1.9 screensaver from Olivier Fabre. I have to run Pointer Prefs and click use setting to get the pointer back. I did not notice this problem under WinUAE version 5.0 and lower.

Additionally, I noticed that when listening to CDDA music using PlayCD, from the mounted BIN/CUE image, there is silence for a second when I switch screens. I have not encountered such a situation in previous versions.

All of the above applies to Beta 2. I haven't checked Beta 1.

RTG or not RTG mode? (It is always important bit)

Switching screen as in switching Amiga screens? Native to different native mode? (different as in PAL to non-PAL or vice versa) Or RTG?

In other words: does it require winuae side changes when Amiga side screen gets changed (different mode = things needs to recalculated etc)

[Ponki1986]

For now I only noticed for RTG modes.

[Waccoon]

- Allow AUDxPER=1, game Spaceport requires it. Previously min allowed audio period value was 4 to keep CPU usage lower if program keeps playing zero single sample continuously.

I've noticed this causes AIBB 5.x to lock up when you click on the benchmark buttons and it produced its beep sound. The program stops responding but multitasking still works. Audio channel DMA gets stuck on with period 1 and a continuous tone is produced. In "audio.cpp" I set PERIOD_MIN back to 4 and it fixes the problem, so it seems to be a special case when the period is exactly 1.

Is the audio system not returning an interrupt when the period is 1, so the program hangs? I haven't gotten around to understanding much about Paula works.

[Toni Wilen]

Originally Posted by Waccoon

I've noticed this causes AIBB 5.x to lock up when you click on the benchmark buttons and it produced its beep sound. The program stops responding but multitasking still works. Audio channel DMA gets stuck on with period 1 and a continuous tone is produced. In "audio.cpp" I set PERIOD_MIN back to 4 and it fixes the problem, so it seems to be a special case when the period is exactly 1.

Is the audio system not returning an interrupt when the period is 1, so the program hangs? I haven't gotten around to understanding much about Paula works.

I couldn't duplicate this and I think AIBB uses audio.device (maybe it depends on KS version) but possible fix is included with winuae.7z. Probably needed very fast CPU mode or something to trigger this.

[Waccoon]

I think this might just be an AIBB thing and it's doing something stupid. Everything else related to audio I've tested, including poking around with the audio registers directly in normal and productivity screen modes, seems to work fine. If another programs locks the audio channels, AIBB plays no sounds and has no problems.

The only lead I have at the moment is that AIBB locks up only if memory cycle-exact mode is enabled. CPU choice, full cycle-exact, or CPU compatibility mode don't matter. Again, Workbench and all multitasking works fine.

When the program makes its beep sound, it locks up with the tone playing continuously in the background. If I manually poke AUD0PER ($dff0ac) to a value other than $1, the audio stops, and the program recovers and continues normally. I have no idea why it even wants a period delay of $1 in the first place.

My debugging skills aren't great, but I'll try to track down what calls AIBB is making.

[Toni Wilen]

Yes but I secretly asked if https://www.winuae.net/files/b/winuae.7z fixes it

[Waccoon]

Yes but I secretly asked if winuae.7z fixes it

Oh, sorry. Nope, still the same behavior.

It seems that if the audio period is too fast in cycle-exact mode, either int 4 isn't being generated correctly or if it is, the OS isn't seeing it. The faster the period, the less reliably is works (setting it to $2 does actually fix it, but only sometimes. $4 always works). I've never done anything with audio before, so I'm not sure where to look for more clues.

[Toni Wilen]

Upload your aibb executable, thanks (aibb shouldn't need anything else?)

[Waccoon]

Here's my AIBB copy (6.5)

AIBB v6.1 has the same behavior. Both versions have the same issue on OS1.2, 1.3, and 3.1.

[Toni Wilen]

Originally Posted by Waccoon

Here's my AIBB copy (6.5)

AIBB v6.1 has the same behavior. Both versions have the same issue on OS1.2, 1.3, and 3.1.

I still can't duplicate it = include config file. (I assume it happens even when your config is booted without startup-sequence)

[Waccoon]

Aha! I'm a USA guy, so while in Workbench I always use my system in NTSC mode and completely forget about PAL. Turns out, PAL mode works perfectly, it's only NTSC mode that has the problem.

For all testing, I was using stock NTSC Quickstart configs with "best compatibility". If you still want to try the config I normally use, it's set up like my real A1200, with no extensions or RTG stuff.

[Toni Wilen]

Originally Posted by Waccoon

NTSC Quickstart[/url]

That was the important bit. Fixed.

[Genju]

Not sure if anybody has mentioned this yet, but starting from beta 1, the pixel shader I've been using (and a few others) are broken.



The shader I'm using is "D3D: 5xBR-v3.8c"

One of the shaders that still seems to work is "D3D: CRT.D3D".

[Toni Wilen]

Originally Posted by Genju

Not sure if anybody has mentioned this yet, but starting from beta 1, the pixel shader I've been using (and a few others) are broken.



The shader I'm using is "D3D: 5xBR-v3.8c"

One of the shaders that still seems to work is "D3D: CRT.D3D".

I don't think it is shader related because there are no changes in shader related code. Perhaps it is shres on hires or hires on lores? It can cause missing horizontal pixels. (and this can be different depending on used config)

[Genju]

Originally Posted by Toni Wilen

I don't think it is shader related because there are no changes in shader related code. Perhaps it is shres on hires or hires on lores? It can cause missing horizontal pixels. (and this can be different depending on used config)

Any ideas on what I could try changing in my configuration?
Resolution is set to Hires (normal), changing it to Superhires doesn't fix it. Line mode is set to Double (Single still results in the glitches in the screenshot).

When rolling back to Winuae 5.0.0. everything looks fine again.

Edit: You might be right, the problem doesn't seem to be the shader after all. I had set it to "Null" for RTG modes and my WB setup is using RTG.
Edit 2: Changing the color depth in the Screenmode prefs fixes it temporarily. Until I reboot the system, at least. Then it's back to looking glitched.