WinUAE 4.10.1 beta series

[tygre]

Originally Posted by Toni Wilen

Assuming you mean that short line in arrow image? (Always include some details, especially when images are blurry..)

Does it happen in 4.9.1 (or older)? (4.10 added more Picasso96 blitter emulation functions)
Does it still happen if you set NOBLITTER=yes (uaegfx.info settings)

In 4.9.1 and 4.10 (and 4.10.1b1 IIRC), there would be a Guru. In 4.10.2 (41010b2.7z), there's only this graphic corruption in two places: in the top icon bar of Cubic IDE and "behind" Cubic IDE, a grey vertical line 25px wide, starting at 685px.

I'm using the HighGFX monitor without any graphic or RTG boards, like on my real Amiga

[Toni Wilen]

Originally Posted by tygre

In 4.9.1 and 4.10 (and 4.10.1b1 IIRC), there would be a Guru. In 4.10.2 (41010b2.7z), there's only this graphic corruption in two places: in the top icon bar of Cubic IDE and "behind" Cubic IDE, a grey vertical line 25px wide, starting at 685px.

If that happens when it crashes/hangs, it probably is side-effect of it (memory corruption).

Enable on screen leds to see if emulation has stopped (HALT or H)

I'm using the HighGFX monitor without any graphic or RTG boards, like on my real Amiga

You should mention it (UAE + "productivity" software + high resolution native mode is rare). Or have OSD enabled in screenshots.

Anyway, I don't see any emulation problems (except that log message which I hopefully solved already).

[Toni Wilen]

https://download.abime.net/winuae/fi...uae_41010b3.7z
https://download.abime.net/winuae/fi...e64_41010b3.7z

Beta 3: (4.10.1 final in about 1 week or so)

- Serial port RI (ring indicator) was inverted. If both printer and serial port was enabled, printer port SEL was pulled down even when RI was inactive.
- DSKBYTR behavior updated after more testing. It is updated all the time normally except if DSKLEN WRITE bit is set without DMA enable. It is even updated (with zeros) when disk write DMA is currently active.
- Dumping CPU state after JIT unhandled exception or program jumping to IO space won't open console window anymore.
- If 2 back to back internal cycle count events had same cycle count value, order of executed events was undefined. Now they are executed in FIFO order. Interrupt clearing and setting at same cycle queued two events, interrupt was lost if they were executed in opposite order. Fixed weird music issue in Jesus on E's / LSD.
- Fixed PCI bridgeboard crash if non-existing PCI board IO or RAM space was accessed.
- FM801 does not seem to have PCI Device control register Interrupt Disable bit (marked as reserved in datasheet). Made it unchangeable (and few other PCI config bits) to fix Mediator FM801 driver audio playback hang. (Possibly this worked with some older Mediator drivers)
- Added command line option "-winekeyboard" to change handling of PAUSE and NUMLOCK keys as a workaround for different (buggy?) Wine RawInput key codes and flags. (falcon35180)

[Ponki1986]

Originally Posted by Toni Wilen

https://download.abime.net/winuae/fi...uae_41010b3.7z
https://download.abime.net/winuae/fi...e64_41010b3.7z

Beta 3: (4.10.1 final in about 1 week or so)

- Serial port RI (ring indicator) was inverted. If both printer and serial port was enabled, printer port SEL was pulled down even when RI was inactive.
- DSKBYTR behavior updated after more testing. It is updated all the time normally except if DSKLEN WRITE bit is set without DMA enable. It is even updated (with zeros) when disk write DMA is currently active.
- Dumping CPU state after JIT unhandled exception or program jumping to IO space won't open console window anymore.
- If 2 back to back internal cycle count events had same cycle count value, order of executed events was undefined. Now they are executed in FIFO order. Interrupt clearing and setting at same cycle queued two events, interrupt was lost if they were executed in opposite order. Fixed weird music issue in Jesus on E's / LSD.
- Fixed PCI bridgeboard crash if non-existing PCI board IO or RAM space was accessed.
- FM801 does not seem to have PCI Device control register Interrupt Disable bit (marked as reserved in datasheet). Made it unchangeable (and few other PCI config bits) to fix Mediator FM801 driver audio playback hang. (Possibly this worked with some older Mediator drivers)
- Added command line option "-winekeyboard" to change handling of PAUSE and NUMLOCK keys as a workaround for different (buggy?) Wine RawInput key codes and flags. (falcon35180)

With reference to this thread performance has increased a bit. Timedemo in Quake gave an average score of 22.6 FPS. Version 4.10 gave a score of 21.7. The average SysInfo score was 396.75 MIPS, AIBB gave a score of 127288.33 Dhrystones for the 68000 (???) code. On the other hand, for the 68020 code the result was 380592.81. Other features like FloppyBridge, PiP, MIDI playback, D3D 9 and 11 and using a PNG image as an Overlay work well.

[a/b]

callm modes with extra operand words are not disassembled properly: second opcode word (immed. byte operand) is skipped too late.
disasm.cpu, lines 2205 and 2206 should be switched:
====
pc = ShowEA(NULL, pc, opcode, dp->sreg, dp->smode, dp->size, instrname, &seaddr2, &actualea_src, safemode);
add_disasm_word(&pc, &bufpc, &bufpcsize, 2);
====

[White]

Toni,
I report a crash .dump (OS4)
I am also attaching the new config file
completely new dedicated only to "Voodoo3"
pertains to betas 1-2-3
the circumstances seem random
in this case the .dump is from beta 3

the crash occurs in case of restart
i had thought about when i add of a directory

but instead it also happens in other cases and seems random

a "blue" screen appears and then generates the .dump file

[Ponki1986]

Is it possible to add a USS file saving option to WinUAE using UAE-CONFIGURATION in CLI/Shell?

Normally you can load such a save state using the syntax without any problem:

uae-configuration statefile .\SaveStates\mystatefile.uss

Saving, on the other hand, could look like this:

uae-configuration savestatefile .\SaveStates\mystatefile.uss

[White]

Toni,

I also leave this .dump (( OS4 + Voodoo3 ))
this is the first time
that the crash occurs without the restart.
but simply by launching winuae the first time.

[Toni Wilen]

Originally Posted by White

Toni,

I also leave this .dump (( OS4 + Voodoo3 ))
this is the first time
that the crash occurs without the restart.
but simply by launching winuae the first time.

This is strange because it crashes when trying to draw chipset sprites that should not ever be visible during boot to OS4..

When exactly does it crash? Include winuaelog.txt with -logflush command line parameter and config file.

[White]

Toni,

the crash occurs at boot a "blue" screen remains
also i noticed that in the latest versions some times (OS4)
on boot before you see the "pink" color always happens on boot
a the screen before turning "pink" you see practically all the "rainbow" colors on the screen
In previous versions this never happened.
You could only see the "pink" color on boot and not all colors.

[Toni Wilen]

Wrong log file. And -logflush is missing.

[White]

how do I run -logflush to enable full logging ?

it won't let me add the -logflush parameter

thank you.

[Toni Wilen]

Easiest is to open windows command prompt ("cmd.exe") and cd to WinUAE directory and type "winuae -log -logflush".

Or add -log -logflush at the end of WinUAE shortcut icon "Target" field.

[White]

Crashed on first boot without restart
Better yet it is probably more accurate.

I hope that's right.
Thank you

[Toni Wilen]

Still wrong log. winuaelog.txt. (NOT winuaebootlog)
EDIT: make sure your config has misc panel winuaelog option checked.

[White]

I hope that's right :-)

[Toni Wilen]

That log was good

Color register writes simply happen too quickly now, overflowing one buffer.. (In accurare modes these are limited to one/DMA clock cycle but fast CPU modes allow it to happen as fast as possible and sometimes it can happen really too fast)

Should be fixed in next beta.

btw, enable at least one floppy drive to speed up booting. Missing DF0: can make booting noticeably slower.

[White]

Toni,

Thanks again for your work and patience :-)

yes, you had already explained to me that the boot is faster with the floppy.
but out of laziness I always leave it deactivated :-)

[Toni Wilen]

https://download.abime.net/winuae/fi...uae_41010b4.7z
https://download.abime.net/winuae/fi...e64_41010b4.7z

Beta 4: (more old bugs found and fixed, at least 1 week more before official release)

- Added continuous screenshot mode to GUI (Was previously available via input events). When ticked, screenshot is taken automatically every frame.
- Fixed video recording and continuous screenshot including extra blank frame when loading new statefile. (4.9.0)
- Fixed second frame appearing corrupted after loading statefile due to not re-initializing long/short frame internal variables at startup (only visible in video recording or screenshot). (4.9.0)
- If game controller was configured using Game Ports panel (Custom #x setting) and device didn't have any Input panel mappings, config file input device matching with currently installed input devices was not done. This meant if config was moved to other PC, device specific input config was not loaded correctly. Now it also matches devices using only USB VID and PID if better match is not found. (Full match -> Partial match -> VID+PID only if exists -> Device friendly name)
- If config file has input device that is not currently connected, saving the config file didn't save non-connected device's config data. Now it is saved (like it was supposed to already do)
- Game Ports panel custom settings were not corrected automatically if device indexes changed (new input devices connected or removed)
- If Game Ports custom setting has entries that don't map to any currently connected device, show unmapped events as "<event name>=?" instead of showing nothing.
- CALLM disassembly fixed.
- 68030 MMU state was partially reset when loading statefile.
- If fast CPU mode program writes to color registers (usually COLOR0 to create background color rainbows), skip write if previous write already happened in same chipset cycle. This could have overflowed buffer that stores state where and when color change should happen in rendering part of emulation. Fast CPU modes don't slow down CPU by emulating chipset or chip ram delays.
- Reduced size of sprite buffers (which was huge), buffer size was enough to have whole screen filled with sprites in max superhires/800 lines screen mode which can't happen.
- "uae-configuration statefile_save <filename>" can be used to create statefile (note that save happens when vblank starts next, it is not saved immediately but uae-configuration still returns immediately)

[hexaae]

b4:
- a few flickering pixels with Shadow of the Beast WHDLoad in main menu: parallax scrolling mountains in background and "Beast BY Reflections" logo in front.