WinUAE 3.4.0 beta series

[Jess2]

Big problem since the anniversary update from micrososoft.

I have win10, nvidia, doublebuffering.

The f12-quit key not working anymore in ocs-ecs-aga-fullscreennmodus.

The monitor is black no input is possible.

Must r e b o o t.

In window mode its ok.

[vagrant]

Originally Posted by Toni Wilen

Ok, so it can also happen when buffer is too small.

Buffer size calculation in wasapi has changed, small buffer is now really small. (pull mode can handle extremely small buffers without glitches, push could not)



wasapi pull mode is not yet really tuned for low latency vsync.

Ok great!
How many samples is "1" for sound buffer now? (or is it not handled like that anymore). Some of my older configs have max_buffer set to 256, which was the lowest my machine could take without noticeable problem.. but of course any audio issues I've reported in the past were using a "normal" config

[Toni Wilen]

Originally Posted by Jess2

Big problem since the anniversary update from micrososoft.

I have win10, nvidia, doublebuffering.

The f12-quit key not working anymore in ocs-ecs-aga-fullscreennmodus.

The monitor is black no input is possible.

Must r e b o o t.

In window mode its ok.

I assume this also happens with previous winuae version (3.3.0)? Because this thread is only for 3.4 problems.

Do you also mean ALT+F4 won't quit it? Or ALT-TAB does not work? (Which means OS or driver bug)

[hexaae]

@ALL
Be sure you're NOT USING nVidia drivers 372.54 because they have a number of known issues with many games too...
Vsync, g-sync, textures, strange crashes, shadows are also affected.

https://forums.geforce.com/default/t...ased-8-15-16-/

368.81 WHQL or .95 (hotfix) were ok.

[Toni Wilen]

Originally Posted by ED-209

Ok great!
How many samples is "1" for sound buffer now? (or is it not handled like that anymore). Some of my older configs have max_buffer set to 256, which was the lowest my machine could take without noticeable problem.. but of course any audio issues I've reported in the past were using a "normal" config

256 samples. (It is in the log). Old was 2x or 4x larger.

Note that buffer config entry does not match actual buffer, it is just artifact of ancient times..

[Toni Wilen]

Originally Posted by mark_k

I tested the attached config with 3.4.0b2. A4000/040 + Picasso boots OpenBSD. You hopefully have the corresponding HDF for that already, though it may not matter since the problem can be seen at the initial CLI prompt.

In that config, the CPU Idle slider is all the way to the right. Start emulation, type loadbsd at prompt. After a while the CPU% indicator says 0% for much of the time. (In fact it shows 0% just sitting at the CLI prompt.)

Open settings window and move the CPU Idle slider all the way to the left. Continue emulation and CPU% still shows as 0% for much of the time.

In comparison, WinUAE 3.3.0 behaves as I'd expect; CPU% is 18 to 21 at the CLI prompt with CPU Idle all the way to the right, 100% with CPU Idle all the way to the left.

CPU Idle changed completely, I just forgot to mention it in change log.

Now it is just two states, active or inactive. Slider selects how idle Amiga needs to be before state changes. Active = fastest possible CPU mode, inactive = temporary internal switch to approximate cpu mode. This guarantees very low CPU usage when Amiga is idle without breaking compatibility.

It is possible DMA HD transfer may not yet cause idle state change properly.

[Toni Wilen]

Originally Posted by michaljarz

I also noticed a problem with sound. As an example, I tested demo "SPACEBALLS - State Of The Art". Config based from quickstart - A500, 1MB Chip, 68000, 24bit adressing, CPU Frec 2x, ECS Agnus, Cycle-Exact. I tested in window mode, but in full screen is same effects.

Windows XP SP3.
Sound driver - DirectSound.

In latest winuae.7z only "Legacy VSync" works OK.
Attached comparision table and logs from 3.3 and latest 7z.

(EDIT: In table => LegacyVSync+DirectDraw = +23% / +25% sound)

Does it also work same as 3.3.0 in non-vsync modes? It is only low latency vsync that works strangely?

[Toni Wilen]

I think I have found the reason for low latency vsync slowdowns (which can also affect other modes). No fix yet..

EDIT: may be fixed (at least partially) in winuae.7z

[michaljarz]

Now it works much better (non vsync and low latency vsync). In DirectDraw mode it is probably the same now as in 3.3 (I think so), Direct3D is a little slower than 3.3 but much faster than it was before (3.4.0 b1 & b2). Thanks! You are great!

PS. I was just during the preparation of films to show the differences between 3.3 and 3.4b. Altogether now, they probably are not needed now:
(video recording with phone )
If littering is that I delete videos from the forum.

3.3.0 without vsync - [ Show youtube player ]
3.3.0 low latency vsync - [ Show youtube player ]
3.4.0b2 without vsync - [ Show youtube player ]
3.4.0b2 low latency vsync - [ Show youtube player ]

[Retro-Nerd]

Indeed. The Low Latency Vsync problem is fixed now. Great!

[michaljarz]

Now I've discovered something new. In the current beta, my system does not start when I turn on "RTG board => Hardware vertical blank interrupt". In 3.3.0 it was OK.

[Snake79]

Using latest winuae.7z with my config (with Fullscreen Low Latency VSync also) the speed / reaction of OS41FE is back to normal . I still have freezes with it (f12, alt-tab, ctrl-alt-del, end+f9 isn't working and i have to restart PC then) after the boot sound (without cracking), but found the reason of it: it was SB128 that i was using (that's working without issues with 3.3.0). After disabling SB128 issue is gone and everything is working as it should (with Paula).
EDIT: attached config have SB128 disabled already.

[mritter0]

The newest build didn't fix the SB128 problem. New logs attached.

The only thing I changed is I installed the DiskImage.device package. I will check in to that.

[Toni Wilen]

RTG hardware interrupt and SB128 hangs are now fixed.

[michaljarz]

Thanks! RTG VB interrupt works!

[michaljarz]

Since 3.4.0b Wazp3D doesn't work - crash when run Cow3D.
3.3.0 and previous work OK.

[Toni Wilen]

Originally Posted by michaljarz

Since 3.4.0b Wazp3D doesn't work - crash when run Cow3D.
3.3.0 and previous work OK.

The usual answer: crash dumps from non-official betas are useless. It also crashes somewhere inside soft3d.dll, original reason is unknown.
(I also don't like native API, it is mainly ugly hack and does not have any real API design..)

EDIT: need much simpler test case..

[AMIGASYSTEM]

Originally Posted by michaljarz

Since 3.4.0b Wazp3D doesn't work - crash when run Cow3D.
3.3.0 and previous work OK.

3.4.0b Wazp3D with OS3 (AFA OS) works well (I attach little video)

[mritter0]

Yes, SB128 is now fixed. Thank you!

[michaljarz]

I tested Wazp3D with many configs - unsuccessfull. 3.3.0 work, 3.4.0 not. Fresh Workbench 3.1 installation also.