WinUAE 2.4.1 beta series

[rsn8887]

Originally Posted by Toni Wilen

It is normal when using OCS/ECS configuration and hires mouse cursor is enabled (prefs/pointer).

I see, my prefs were set to low-res pointer. But it could be MCP or some other patch something messing it up sometimes?! Anyways, it just happened once in 2.4.0, too, so it is OT.

[Toni Wilen]

Originally Posted by rsn8887

I see, my prefs were set to low-res pointer. But it could be MCP or some other patch something messing it up sometimes?! Anyways, it just happened once in 2.4.0, too, so it is OT.

MCP can do lots of things

I am starting to wonder if your problem is actually caused by some patch or something similar because it really does not make any sense to have smooth screen dragging and game scrolling but strange mouse pointer movement.

Fastest possible modes can have wildly different timing depending on used PC speed, vsync settings, sound settings etc..

Do you get that strange effect if you boot without startup-sequence? Also test KS 1.3 too.

[Toni Wilen]

and another winuae.zip update.

Now fastest possible should be more like 2.3.3 behavior, extra CPU emulation is run at least every 8th line (like 2.3.3 and older did in JIT mode), it also still runs after every line if there is enough time left (like in 2.4.0+), it shouldn't break 2.4.0+ performance increase.

Previously CPU time may have been distributed very irregulalarly which could have caused reported side-effects.

[Toni Wilen]

http://www.winuae.net/files/b/winuae_2410b10.zip

Beta 10:

- Partial rewrite of fastest possible CPU low latency vsync detection code. Should fix jitter caused by frame not always getting shown on screen at correct time.
- Low latency vsync and double frame mode (85Hz+): guarantee every other vblank is really skipped, previously it didn't work correctly in some situations.
- Calculate final vblank busywait time from selected rate instead of using hardcoded 50Hz/100Hz-only compatible value. Fixes >50Hz no-buffer jitters.
- Emulate extra CPU time if it has been more than 7 lines since last extra CPU time, matches older behavior without losing 2.4.0+ performance increase.
- Fixed low latency vsync + fastest possible CPU getting in state where FPS dropped greatly temporarily, usually at startup.
- Show also drives that require admin privileges in harddrives panel, unselectable and marked with "Access denied" label. Only name shown because accessing any other attribute may require admin privileges.

[Leandro Jardim]

A small line appears when I use a image mask on my Radeon HD4550 with Beta 10. In full screen it appears on top whereas on window it appears on bottom.

[rsn8887]

The mouse pointer still not smooth for me in b10.

Toni, to answer your question, I am not using any patches when testing this. Just quickstart A1200 + original Workbench 3.0 disk (and fullscreen mode btw).

It behaves as if the "fastest possible" emulation is not "leaving enough time" for the mouse pointer motion. Maybe in the old days option "fastest possible but maintain chipset timing," chipset timing included timing allocated for mouse pointer smoothness, and now it doesn't anymore?

If I reduce the emulated CPU speed enough (-70% or -80%), the pointer seems to get "enough time to be perfectly smooth". If I reduce the emulated CPU by only 20%, the pointer gets smoother but not perfectly smooth yet. For some reasons, in latest beta -10% makes everything horribly slow for me (fps display stays at 60, but my screen refreshes only once every second or so).

If I boot into my own WB with a lot of bells and whistles, patches etc, the pointer is more jerky than in an original WB where "not much is going on."

If I boot into my own WB, and move the screen all the way to the bottom, so that only a few lines of pixels are displayed, the mouse pointer is already almost perfectly smooth at only -20% CPU! Still jitters though This is in 100 Hz mode. If I drag the screen back up, mouse pointer becomes more jerky.

Here's a baffling result to me: I boot into original WB3.0, in fastest possible A1200 low latency vsync, this is in 60 Hz mode. Mouse pointer is not smooth, or better said, a "ghost image" flickers every second or so behind mouse pointer. Then I change from line mode "double" into "single", where the displayed area in fullscreen is only half as big, and the mouse pointer becomes almost perfectly smooth! Only one jitter every five or so seconds! Then I change vertical scaling to 2* and pointer stays this smooth!

My theory is that line doubling mode is "slower" than the combination of single line mode and 2* vertical scaling, and somehow this combination allocates more available time to the mouse pointer motion which is therefore rendered more smoothly.

However, if I boot into my own WB with all bells and whistles, the above trick doesn't make mouse pointer smooth, I suppose there's too much going own (copper bg, eyes and realtime clock etc).

Is there something that was done so that the old "maintain chipset timing" is not really maintained anymore when it comes to the mouse pointer?

[rsn8887]

Originally Posted by Toni Wilen

MCP can do lots of things

I am starting to wonder if your problem is actually caused by some patch or something similar because it really does not make any sense to have smooth screen dragging and game scrolling but strange mouse pointer movement.

Do you get that strange effect if you boot without startup-sequence? Also test KS 1.3 too.

I get it if I boot my own WB without startup-sequence and also when I boot the standard WB3.0 disk that shipped with the Amiga.

It also happens using Amiga 500 config with standard WB 1.3 disk (in fastest possible).

The screen drags smoothly, and the mouse pointer jumps up and down. Same when dragging or resizeing a window. The frame moves smoothly. The point that is dragged moves smoothly, only the mouse pointer jumps back and forth around it. It is as if the pointer temporarily detaches from the dragging point. As if the pointer is just drawn at the wrong position once every few frames, while everything else behaves as if it was actually positioned where it should be if it was moving perfectly smoothly.

EDIT: it also happens in low-res screenmode (thought for a second that it was due to combination of high-res mode and low res pointer but no)

EDIT: tested also in windowed mode, same behaviour.

[andreas]

just jumping in, but just for one thought...

Originally Posted by Toni Wilen

Seriously, I'll do this as long as it is interesting, money won't buy real interest anyway.. I think the only thing that requires some kind of bounty or something is: CD32 FMV cartridge emulation. It is mostly useless and it isn't trivial at all.

Nor is A3000's MMU. (!= 68040/060 MMU, to the others!)
But technically interesting for sure.

And end-of-OT...

-

Toni, with astonishment, I've also read about your HISTORICAL removal of CPU->Chipset slider stuff (yes, historical, because even the old 0.6.9 command-line-only win32 build from *hold your breath* 1996 already had this "feature"!).
May I ask whether this new implementation (using relative values) can also work reliably as a sort of "slowdown hardware emulation" with games?
(What I'm pointing to is those hardware devices actually sold in times of yore, designed for real Amigas and sporting a knob that looked like a volume knob on a Hi-Fi with which you could PHYSICALLY slow down the Amiga CPU, to master games that were much too fast for many folks (e. g. R-Type II))

[Toni Wilen]

Originally Posted by andreas

Nor is A3000's MMU. (!= 68040/060 MMU, to the others!)
But technically interesting for sure.

68030 MMU isn't interesting. It is just implementing fully documented piece of hardware in emulation.

May I ask whether this new implementation (using relative values) can also work reliably as a sort of "slowdown hardware emulation" with games?
(What I'm pointing to is those hardware devices actually sold in times of yore, designed for real Amigas and sporting a knob that looked like a volume knob on a Hi-Fi with which you could PHYSICALLY slow down the Amiga CPU, to master games that were much too fast for many folks (e. g. R-Type II))

It does same if using approximate CPU modes. (CPU frequency setting does the same in cycle exact modes)

It sort of does same in fastest possible modes but because it is relative to current free host CPU time, CPU speed is quite variable.

[Toni Wilen]

Originally Posted by Leandro Jardim

A small line appears when I use a image mask on my Radeon HD4550 with Beta 10. In full screen it appears on top whereas on window it appears on bottom.

Attach your config file, thanks.

[Leandro Jardim]

I did upload the files.

[rsn8887]

Side note: Thank you Toni for making WinUAE better and better, I agree that the removal of CPU/Chipset slider is absolutely a milestone.

I just noticed the following sound related issue. Tested in A1200 fastest possible llvsync, a bit too tired to test more right now. I am using Portaudio/ASIO.

In latest beta, I get sound buffer underruns: snd LED at -99 and blinks red for a short time (<second) and sound crackles. This happens a bit randomly but every few minutes. In addition, there's in general quite a big range on my snd led (goes anywhere from -99 to +80 or so). Tried so far to
- increase sound buffer
- change priority
- change CPU idle
- change CPU %
but sound buffer underruns remain. Granted, it sometimes takes a few minutes for the first underrun to show up (snd led, after a few minutes, kind of makes it's way to -99, stays there for a while then a red underrun blink!)

In 1.4.1b6 and 1.4.0 release, this does not seem to happen (snd led also goes to -40 or so sometimes, but quickly, after just a second or so, jumps back to small numbers and is mostly around +/- 20).

Tested just letting my WB sit there and watching snd led, but also verified sound crackles during underruns by playing a mod using eagle player.

Logs attached.

[Toni Wilen]

Originally Posted by rsn8887

In 1.4.1b6 and 1.4.0 release, this does not seem to happen (snd led also goes to -40 or so sometimes, but quickly, after just a second or so, jumps back to small numbers and is mostly around +/- 20).

Do you mean b7 is first to have this problem or some later version? Very important to know exact version.

[rsn8887]

Originally Posted by Toni Wilen

Do you mean b7 is first to have this problem or some later version? Very important to know exact version.

The problem occurs only in b8 and later.

[Toni Wilen]

Originally Posted by Leandro Jardim

I did upload the files.

Did you forgot to test older version(s) because I can see it in 2.3.3 too..
(EDIT: and no idea why it happens when other masks work fine..)

[thomas]

"This is unstable beta software. Click cancel if you are not comfortable using software that is incomplete and can have serious programming errors."

Could you move the remembering of this into winuae.ini rather than the registry so that it is shown only once also if you don't have admin rights?

[Toni Wilen]

Originally Posted by thomas

"This is unstable beta software. Click cancel if you are not comfortable using software that is incomplete and can have serious programming errors."

Could you move the remembering of this into winuae.ini rather than the registry so that it is shown only once also if you don't have admin rights?

It does not require admin rights. It is always asked again if exe is modified, even if version is exact same.
EDIT: It also reappear after 7 days.
EDIT2: I meant it shouldn't require admin right..

[Toni Wilen]

http://www.winuae.net/files/b/winuae_2410b11.zip

Beta 11:

- Adjusted Fastest Possible timing a bit more..
- Better PC interlace screen mode detection.
- Fixed invalid MFM encoding when converting standard adf unencoded data to MFM encoding. (Few incorrect MFM clock bits) Another "no one cares" fix (since the beginning, does not affect adf data quality), noticed when testing m68k AROS trackdisk.device write support.
- Vsync mode sound sync shouldn't be that aggressive. (b8)

[Leandro Jardim]

Originally Posted by Toni Wilen

Did you forgot to test older version(s) because I can see it in 2.3.3 too..
(EDIT: and no idea why it happens when other masks work fine..)

Funny... Here I dont see anything wrong in 2.3.3.

[Retro-Nerd]

Something is now completely wrong with the USB joypad, keyboard, mouse detection. There are double dections of my pads/sticks, can't set the mouse in port 1.

Edit: Had to reboot my PC. Now Beta 11 works as it should. Strange.

- Better PC interlace screen mode detection.

Looks great now. Just tried it with some interlaced title screens, e.g. Giana Sisters and Unreal.