WinUAE 4.9.0 beta series (Was 4.5.0)

[Mad-Matt]

Is G-Rex supported for testing in AmigaOS3 yet? (no issues at all in OS4)

As it stands enabling grex + voodoo results in a doubled screen res that doesn't line up with the buttons lol. Maybe something to do with built in mode promotion of the ppc cards that cant be disabled in later firmware.

The effect goes right through to workbench and selecting a voodoo screenmode somehow makes it look worse.

[falken]

Originally Posted by DamienD

Thought I'd better test the latest beta, as I usually do, with all the usual 30 x games that have given me slight issues over the years.

Happy to say that all were fine

...except Aladdin [AGA].

You now no longer see these screens, and just black:

Also, had some slight cracks / pops with some games, even though CPU usage was no where near 100%. Two examples are Agony [AGA] and Disposible Hero.

What filter are you using?

[Toni Wilen]

Originally Posted by thomas

A config with Voodoo gfx card which worked with beta 13 now gurus 0100000C and hangs at black screen.

Mediator voodoomemos is disabled? Mediator for some reason really wants to enable it by default..

Which mediator driver version? (latest MM update?) Include whole medconfig report too. There are far too many variables in mediator configs..

[Toni Wilen]

Originally Posted by Mad-Matt

Is G-Rex supported for testing in AmigaOS3 yet? (no issues at all in OS4)

As it stands enabling grex + voodoo results in a doubled screen res that doesn't line up with the buttons lol. Maybe something to do with built in mode promotion of the ppc cards that cant be disabled in later firmware.

The effect goes right through to workbench and selecting a voodoo screenmode somehow makes it look worse.

CSPPC or BPPC? Check your CSPPC boot menu options. (ESC pressed when booting).

I suspect Voodoo does not have anything to do with the boot menu anyway, it is shown by native chipset.

[DamienD]

Originally Posted by falken

What filter are you using?

None, as my poor old laptop cannot handle filters / shaders well.

All I use is scanlines set at either 50% or 25% opacilty.

Set this under Filters --> Extra Settings --> Scanline opacity.

Whatever you do, do not use Scanlines under Display; they are dark and don't scale to different resolutions well i.e uneven.

[DamienD]

Originally Posted by DamienD

Also, had some slight cracks / pops with some games, even though CPU usage was no where near 100%. Two examples are Agony [AGA] and Disposible Hero.

Hi Toni,

I've been using v4.5.0 betas more and more.

I'm finding the above is happening quite frequently now with various games at various spots. I guess some sound settings changed in these betas, my guess would be the thread for hexaae where he mentioned sound issues so something was changed i.e: WinUAE unstable framerate since 3.4.0b1

Not sure what you want me to do in order to help, logs etc?

[falken]

Originally Posted by DamienD

None, as my poor old laptop cannot handle filters / shaders well.

All I use is scanlines set at either 50% or 25% opacilty.

Set this under Filters --> Extra Settings --> Scanline opacity.

Whatever you do, do not use Scanlines under Display; they are dark and don't scale to different resolutions well i.e uneven.

Thank you mate!

[Mad-Matt]

Originally Posted by Toni Wilen

CSPPC or BPPC? Check your CSPPC boot menu options. (ESC pressed when booting).

I suspect Voodoo does not have anything to do with the boot menu anyway, it is shown by native chipset.

I'm testing with CSPPC atm. The CS Firmware copies the native screen to the pci gfx card which automatically becomes the front screen. (Can switch with end+f9).

EDIT: Quick test with BPPC+Grex+Voodoo, the same issue applies.

[mark_k]

Originally Posted by Toni Wilen

- FAS246 SCSI chip apparently has Features Enable always set (or bit does not exist anymore). DKB RapidFire tests if transfer count high register works (write something, read it back) and assumes it is enabled without modifying Configuration 2 Register. Rapidfire worked when it first implemented because transfer count high was not originally correctly conditionally emulated. Datasheet seems to be MIA.

Which register bit# is the Features Enable bit?

[Toni Wilen]

Originally Posted by DamienD

Hi Toni,

I've been using v4.5.0 betas more and more.

I'm finding the above is happening quite frequently now with various games at various spots. I guess some sound settings changed in these betas, my guess would be the thread for hexaae where he mentioned sound issues so something was changed.

Not sure what you want me to do in order to help, logs etc?

No need do anything yet. I'll finish voodoo and blitter changes first.

Originally Posted by Mad-Matt

I'm testing with CSPPC atm. The CS Firmware copies the native screen to the pci gfx card which automatically becomes the front screen. (Can switch with end+f9).

Oops. I had older CSPPC firmware in my G-REX config that didn't support PCI. Now I can duplicate it.

Originally Posted by mark_k

Which register bit# is the Features Enable bit?

Bit 6, register B. I am quite sure FAS246 (or some other later FAS2xx model) simple dropped the bit and always have extra features enabled.

[mark_k]

Originally Posted by Toni Wilen

Bit 6, register B. I am quite sure FAS246 (or some other later FAS2xx model) simple dropped the bit and always have extra features enabled.

The FAS256 datasheet on Bitsavers (path pdf/emulex/asic/FAS256_Fast-Wide_SCSI_Controller_Jan92.pdf), while not a complete programming manual, shows features enable being bit 4 of configuration register 2.

Maybe FAS246 is a narrow SCSI version of FAS256?

[DamienD]

Originally Posted by Toni Wilen

No need do anything yet. I'll finish voodoo and blitter changes first.

Sure thing Toni, let me know though when you want me to do anything

[Toni Wilen]

Originally Posted by mark_k

The FAS256 datasheet on Bitsavers (path pdf/emulex/asic/FAS256_Fast-Wide_SCSI_Controller_Jan92.pdf), while not a complete programming manual, shows features enable being bit 4 of configuration register 2.

Maybe FAS246 is a narrow SCSI version of FAS256?

Perhaps but config #3 is totally different and read register #9 does not even exists in 53F94. I think FAS256 probably isn't exactly same as FAS246, wide support most likely required even more modifications.

EDIT: This could be even a chip bug, I rechecked what the driver does: If writes to all 3 transfer count registers, then reads then back and checks value matched what was written. _Then_ it sets config #2 bit 6!

[Toni Wilen]

Originally Posted by Mad-Matt

I'm testing with CSPPC atm. The CS Firmware copies the native screen to the pci gfx card which automatically becomes the front screen. (Can switch with end+f9).

EDIT: Quick test with BPPC+Grex+Voodoo, the same issue applies.

I am not sure whose fault this is.. CSPPC firmware sets SVGA attribute register 0x10 bit 6 which doubles pixel width (basically equals Amiga lores mode). But it still sets display width to 640 pixels. So output is technically correct. (minus mismatched mouse position which is a bug)

Perhaps this is some weird (S)VGA feature that is not fully emulated. (EDIT: S3Virge has extra condition when checking this bit so possibly Voodoo also ignores it if some advanced feature is also enabled)

EDIT: Incorrect board type in CSPPC firmware PCI list was a bug but it didn't affect "lores" effect.

[mark_k]

Originally Posted by Toni Wilen

EDIT: This could be even a chip bug, I rechecked what the driver does: If writes to all 3 transfer count registers, then reads then back and checks value matched what was written. _Then_ it sets config #2 bit 6!

Could it be that the transfer length high register can be read/written by the CPU always, but it is just ignored by the chip unless the features enable bit is set?

[White]

Toni
I did the various tests the crash occurs in all cases mentioned Direct3D 9/11 etc. this with the official beta.
While instead with the beta of the day in this case the 2021.01.03
the problem does not occur everything works even in full screen with the same configuration.

[as1mov]

thanks for ALL those fixes Tony, wonderful stuff.

Originally Posted by Toni Wilen

WinUAE 4.5.0 beta series.

As usual, this thread is only for 4.5.0 beta introduced bugs or features. Always test with 4.4.0 first! Problem exists in 4.4.0 or older: do not post in this thread!

https://download.abime.net/winuae/fi...nuae_4500b1.7z
https://download.abime.net/winuae/fi...ae64_4500b1.7z

Beta 1:

- Fully merged latest Aranym JIT. Previously it was only partially merged because of unsolved problem that made it completely unusable. It is solved now.
- JIT should be now trace mode (debugger) compatible. If SR register T1 or T0 bit gets set, JIT transparently drops to normal emulation mode. JIT reactivates when trace gets cleared. Very rarely used M SR bit also does same.
- Full CPU tester 68020 test set now passes in JIT mode, including trace mode tests. (NOTE: bcc.b and bsr.b tests fail unless Hard flush is also set)
- JIT has two known not yet fixed bugs: edge case shift condition (shift count is divisible by 32) C-flag problem and ROR/ROL does not clear V-flag.

- x86 Bridgeboard update: PCem x86 core updated to v16. Update added FPU support for pre-486 CPUs. FPU checkbox added to bridgeboard expansion settings.
- uaegfx Picasso96 2.5.0 screen dragging support added.

- CV64 RTG board emulation. S3 Trio 64 chipset emulation from PCem. Roxxler chip is not (yet?) emulated. CGX4 complains if Roxxler fails to work correctly. Picasso96 does not use it.
- CV64/3D RTG board emulation, Z2 and Z3 mode. S3 Virge chipset emulation from PCem. (S3 Virge is very different than S3 Trio)
- All Cirrus Logic based RTG boards now use PCem Cirrus Logic SVGA chip emulation.
- Piccolo SD64 fixed.
- Reordered RTG board list.
- Hardware emulated graphics board screenshot support fixed.
- x86 Bridgeboard VGA still use QEMU Cirrus Logic emulation. I'll keep it (at least for now) because it enables simultaneous use of bridgeboard VGA and PCem RTG board.

NOTE: JIT + Hardware emulated graphics board can cause glitches. Will be fixed later.

- No. Don't ask. Permedia 2 (Cyber/BlizzardVision PPC) or complete enough Voodoo 3 chip emulation does not currently exist.

- Audio and disk DMA: Agnus also ignores DMACON DMAEN bit. DMAEN bit is checked inside Paula. Audio: Paula audio DMA requests are only generated if DMAEN is set when channel generates DMA request (not when transferred via DMAL). DMAEN can be cleared after request has been generated and DMA request still goes to Agnus and Agnus does DMA to AUDxDAT during channel's DMA slot in next scanline. Disk: both DMAEN and DSKEN needs to be set. Previously DMAEN was checked in Agnus side when deciding current DMA slot state. Most likely no normal program cares, this is quite extreme edge case.
- Copper write to BPLxDAT timing fix.
- cpu_halt_auto_reset config entry added to GUI misc panel.
- Analog joystick multiplier/offset default adjusted.
- Load PAL mode, OCS Agnus state file directly from GUI/command line, start emulation, on the fly switch chipset to ECS Agnus: mode was changed to NTSC.
- Switching off display DMA when display DMA was active still had more hidden side-effects. It results in "bad state" that is not reset until bitplane DMA is switched back on, when it is back on DMA restarts from state which normally happens when DDFSTOP has been passed. Result is either 1 or 2 "bitplane blocks" (1 or 2 depends on horizontal position relative to next BPL1DAT slot when DMA was switched off) starting from DDFSTRT (if switched on before DDFSTRT). The rest of scanline is drawn using border color. Following scan line works normally. No differences between OCS/ECS. (vAmiga test case)
- Another undocumented feature: if BPLCON0 HAM and DPF are both set, result is weird. HAM replaces R/G/B component: replacement component value is palette _index_ of current pixel in dual playfield mode! HAM selection bits zero: palette selection works like in dual playfield mode. BPLCON2 works like mode is normal dual play field (including invalid BPLCON2 priority values). OCS/ECS only. (vAmiga test case).
- Dual playfield mode illegal BPLCON2 priorities didn't work correctly in hires mode. (bottom part of above test case)
- Blitter nasty off cycle counting update, emulatation code wasn't exactly doing what it was supposed to do.
- A1000 blitter busy bug was ignored if copper waited for blitter.
- Added DMA debugger non-nasty blitter extra information: 's' (CPU will get next blitter cycle) and 'S' (CPU stole this cycle from blitter)
- Confirmed CIA register contents at power up. CIA control register should be zero, for some reason previous reset default was $04.
- Disassemble FBF with zero offset as FNOP.
- Debugger fi instruction breakpoint command now accepts up to 3 opcode words.
- Config file list wasn't refreshed when config file was deleted. (4.4.0)
- RTF_AFTERDOS is broken in KS 2.0 , KS 2.0 now uses pre-WinUAE 3.3 hack and only KS 3.0+ use RTF_AFTERDOS. Clipboard sharing and uaenet.device work again under KS 2.0.
- Added Amiga Bootblock Reader brainfile support (copy ABR package xml files to <winuae root>\plugins\ABR -directory). Used when clicking "?" button in quickstart and disk panels.
- Fixed emulated floppy disk rotation speed, it was slightly too slow. This change could break some copy protections. Unexpected side-effects are always possible..
- Copper SKIP wasn't updated when WAIT logic was updated ages ago. SKIP had wrong timing in some rare situations.
- "Default" filter mode: PAL/NTSC switch automatically scales the image.
- DF3: floppy led was drawn using very dark color if it was disabled and system didn't have NVRAM hardware.
- OSD led font graphics is now generated from real Windows font and is high DPI aware. Led size multiplier config entries now also support non-integer values.

- Added GVP A1230 Turbo+ Jaws. Boot ROM adds RAM outside of 24-bit address space and fake RAM autoconfig board.
- Accelerator board RAM slider range was larger than supported range with some boards.
- Added Trumpcard v1.5 and v3.1 ROM. Added also hardware interrupt checkbox, later trumpcard (still non-pro) variants have interrupt support. It seems 8k ROM variants don't have interrupt, 16k do have it. Setting is wrong if it hangs at boot.
- Added A.L.F.2 OMTI variant and v2.2 boot ROM. ROM possibly also includes SCSI driver but I didn't examine how it works. A.L.F.2 OMTI IO port is exact same as in A.L.F.1.
- Trifecta LX SCSI emulation fixed.

[White]

I have noticed that when I use a directory on the desktop to access amiga files for convenience
I cannot increase the Z3 memory the maximum I can use and 256 ram
For example, if I put 512 of Ram in Z3 on the first bank
On reboot it gives me the error in Expansion Board "DEFECTIVE"
If I don't use the directory on the desktop, I can increase the memory without problems.
Do I make any mistakes?
Need a screenshot?
Thanks

separate note:
system in use 4.1FE Voodoo3 Mediator

[amilo3438]

Psylteflesk by Ephidrena -> http://www.pouet.net/prod.php?which=50125
Picnicbass by Ephidrena -> http://ada.untergrund.net/?p=demo&i=671

The problem with the "pre-amplified" sound starting from WinUAE_4500b1! (fine in official WinUAE_4400)

Tested Configuration
-----------------------
WinUAE Quickstart
Model: A4000
Configuration: 68030, 3.1 ROM, 2MB Chip + 8MB fast

Drives:
DF0: "Workbench v3.1 rev 40.42"
DH0: "eph-psylteflesk.lha" (or "eph-picnicbass.lha")

NOTE: One of above demos is attached as archive via "CD & Hard drives / Add Directory or Archive... / Select Archive or Plain File"

[Toni Wilen]

Originally Posted by White

I have noticed that when I use a directory on the desktop to access amiga files for convenience
I cannot increase the Z3 memory the maximum I can use and 256 ram
For example, if I put 512 of Ram in Z3 on the first bank
On reboot it gives me the error in Expansion Board "DEFECTIVE"
If I don't use the directory on the desktop, I can increase the memory without problems.
Do I make any mistakes?
Need a screenshot?
Thanks

separate note:
system in use 4.1FE Voodoo3 Mediator

Usual stuff, config and log + screenshot. (and also screenshot from hardware info panel)

Originally Posted by amilo3438

Psylteflesk by Ephidrena -> http://www.pouet.net/prod.php?which=50125
Picnicbass by Ephidrena -> http://ada.untergrund.net/?p=demo&i=671

Looks like JIT FPU related. (always remember to test with JIT off!) Possibly some edge condition (infinity for example) that just worked accidentally, x86 FPU vs 6888x is far too different.

Debugging later..