WinUAE 3.5.0 beta series (was 3.4.1)

[Toni Wilen]

Originally Posted by bladecgn

On "Chipset Extra" drop-down "A1200" is listed twice.

That is static list, I don't see how it can have duplicates.

[Ian]

A1000
A1200
A2000

Showing for me, maybe a quick glace at them could've made them look like dupes.

[mcbpete]

Originally Posted by DamienD

Heya Toni,

Haven't been that active in testing the 3.4.1 betas so far, tonight I thought I better get more involved

...obviously I'm not about to test all 1690 games / configurations that I have in my WinUAE collection so thought I'd test my usual list of *difficult* games over the years that sometimes don't work when a new version is released:

.... Game list ....

Curious about this ... My configs are system based rather than game based (e.g. I've configs called A500 (512k), A500 (1mb), A500+, A1200, CD32), and then run the games based on what system they're for. Is there a benefit of doing this the other way round ?

[DamienD]

Originally Posted by mcbpete

Curious about this ... My configs are system based rather than game based (e.g. I've configs called A500 (512k), A500 (1mb), A500+, A1200, CD32), and then run the games based on what system they're for. Is there a benefit of doing this the other way round ?

Firstly, sorry Toni for the off topic...

Hmmm, how best to answer this one...


A500:
Most A500 games run with either:

a) OCS or ECS / KS 1.3 / 512KB Chip and 512KB Slow RAM.
b) OCS or ECS / KS 1.3 / 1MB Chip RAM.

Others don't though so may require KS 1.2 or only 512KB Chip RAM for example etc.


A1200:
Most A1200 games run with:

a) AGA / KS 3.0 / 2MB Chip RAM.

Other don't though so may require KS 3.1 or some Fast RAM.


I also usually adjust other things to best suit a specific game e.g:

a) Chipset: "Cycle-exact" is on the majority of the time but sometimes you need this off. Not to mention a few games require "Immediate Blitter" (quite rare though).
b) Display: is either set as full screen 640x480 or 720x576 for my laptop (no external monitor but connect to my TV via HDMI on occasion).
c) Centering: most of the time horizontal / vertical centering works fine but not always (screens will jump into place which I don't like).
d) Positioning: if centering is switched off due to above then sometimes I need to adjust horizontal / vertical positions manually to have the game centered on screen.
e) Joystick: my default setup (when my Xbox 360 controllers aren't connected) is "Keyboard Layout B". With some games you need to change this to be "Keyboard Layout A" as you need the standard cursor keys. Pinball games are a good example of this.
f) The best disks to be used and added into the number of supported drives. Also, for multi-disk games I use of "Disk Swapper" and setup accordingly.


Then of course there are the hard drive / WHDLoad / AmigaCD games which all require different settings i.e. processor / RAM etc...


I'm sure you get all of the above. I've been using WinUAE since 2005 and have always made one configuration file per game. This means that after making all the desired adjustments I know that when I start the game everything is perfect, no messing around

...but it has taken me years to get my collection to how I like it. It's much the same as what GameBase Amiga does; one configuration per game.

You can see my "sort of" current games list in this thread: Request for any gems. Now imagine trying to remember the best files / required settings for 1690+ games?

...which is what others would have to do if they only have configurations setup by model.

[Toni Wilen]

http://www.winuae.net/files/b/winuae_3500b10.7z
http://www.winuae.net/files/b/winuae64_3500b10.7z

Beta 10:

- Version bumped to 3.5.0. .x was only reserved for minor bug fix versions.
- GUI listview column width calculation used default font size, not selected font.
- Z3 autoconfig in Z3 space didn't support documented byte write to 0x44 config register. (KS ROM always uses word write)
- Manual RAM configuration GUI end address validation was incorrect. Now end address can be smaller than total RAM board size.
- Fixed A2090(a) ST-506 emulation, last accessed LBA status field must be always filled if read/write/format command, added disable ST-506 option. (ROM driver checks bit in autoconfig data if ST-506 is available)
- MMU emulation can be now switched on/off on the fly. (Of course AmigaOS will crash if MMU is switched off and MMU is used for Map ROM etc..)
- MMU PTEST in 68040 mode without MMU emulation now checks transparent translation registers and returns correct physical address, now Enforcer initializes correctly in non-MMU mode and does not cause crash due to wrong MMU tables if MMU is later switched on.
- Fixed Blizzard accelerator crash if memory size was zero (don't map empty maprom space)

[DamienD]

Many thanks Toni

Is the full version of 3.5.0 imminent?

Originally Posted by Toni Wilen

- Version bumped to 3.5.0. .x was only reserved for minor bug fix versions.

You should change the thread title then

[ross]

Finally replicable.
Like I've written in some old messages, I sensed of some problems with MMU enabled but subtle and unpredictable.
Now you can view in action.

Use my usual (attached) configuration, but another with MMU enabled can be fine, boot from my NTSC patched version of Darkseed (in The Zone) and the sprite/bpl party begin!
Seems that copperlist do not start / set pointers but is not the case...
If you enable/disable MMU (latest beta) in real time the problems is enabled/disabled in real time

All begin with beta3 (beta2 is working properly).

Cheers,
ross

[Toni Wilen]

It isn't emulation bug. (At least not directly)

Game's copper list and KS ROM vblank code both write to DFF084 and for some really odd reason it just happens that current "68030 cycle exact" (especially in MMU mode it can be strange) KS ROM code manages to do the write exactly before copper executes COPJMP2..

[ross]

Originally Posted by Toni Wilen

It isn't emulation bug. (At least not directly)

Game's copper list and KS ROM vblank code both write to DFF084 and for some really odd reason it just happens that current "68030 cycle exact" (especially in MMU mode it can be strange) KS ROM code manages to do the write exactly before copper executes COPJMP2..

Ok, the patch is simple (avoid using COP2, 'reuse' COPLC1/JMP1).
But the point is the different behaviour in MMU emulation mode .

(I've seen similar oddities in other cases, making difficult the debugging)

Thanks Toni, bye,
ross

[Toni Wilen]

Nothing prevents similar issues to happen with real hardware (or some less real..), it just depends on CPU caches, CPU speed, ram speed, chipset DMA activity etc..

[ross]

This makes the "real-time" MMU enable/disable more meaningful.
This 'slightly different' cycle-exact mode may help catch some nasty bugs

[Toni Wilen]

http://www.winuae.net/files/b/winuae_3500b11.7z
http://www.winuae.net/files/b/winuae64_3500b11.7z

Beta 11: (max 1.5 weeks to go)

- 68060 PLPA MMU instruction now returns data that makes more sense in non-MMU mode. (Similar change as b10 68040 PTEST) (Emulated 68030+ always appears as MMU capable from software point of view for compatibility reasons, even if MMU emulation is disabled)
- Do not enable Logitech LCD support if init succeeds but no displays available.
- Listview font width calculation was still wrong if font was same as GUI font.
- Use softfloat routines (not normal doubles) to output debugger FPU register contents in softfloat mode. (AG)

[Sinphaltimus]

Originally Posted by Toni Wilen

- Do not enable Logitech LCD support if init succeeds but no displays available.

Sweet.

[DamienD]

Originally Posted by Sinphaltimus

Sweet.

...so you're issue is now solved with this latest beta?

[Sinphaltimus]

Originally Posted by DamienD

...so you're issue is now solved with this latest beta?

Yep.

[michaljarz]

Starting with version 3.5.0b3 (3.4.1b3), some configurations have a problem with the switching of some sound settings in the fly. When "Audio filter" is set to "Emulated (A500)" or "Always on (A500)", changing any other sound settings (eg "Stereo separation") causes sound to disappear (only when emulation is running).

[Toni Wilen]

Originally Posted by michaljarz

Starting with version 3.5.0b3 (3.4.1b3), some configurations have a problem with the switching of some sound settings in the fly. When "Audio filter" is set to "Emulated (A500)" or "Always on (A500)", changing any other sound settings (eg "Stereo separation") causes sound to disappear (only when emulation is running).

Can you duplicate it in unexpanded A500 or A1200 configuration? (Less unknown variables).

Does Sound emulation disabled checkbox get checked when it happens?

[michaljarz]

This is happening with a very specific configuration. I created a configuration based on "A1200 non-expanded", then I turned on FPU 68882, added a hard drive (directory), set the audio filter to "Emulated A500" and started emulation. Then I started Octamed Sound Studio (or ProTracker) and loaded the module. During playback when it comes to settings (F12) and changes in options such as "Stereo Separation" to some other value, then when the emulation returns the sound is missing. "Sound Emulation" is still referred to as "Enabled" (disabling and enabling nothing here, you have to restart WinUAE).

In conclusion, this happens only when I enable FPU (but plain, SoftLoad is OK) and change "AudioFilter" to "Emulated A500".

[SaphirJD]

From Beta 8-present there was some kind of bug introduced which screws up demos.

Config is normal A500 quickstart 1.3 OCS config - More details of course in the problematic demo thread. As that was already quite a while since it was introduced, have been thinking it would make sense to report it here too. Hope that was not too wrong

[Toni Wilen]

This is more correct thread (at least in theory..), something that worked previously got broken.

It looks like my blitter BDAT special case tests were not good enough. Back to testing