WinUAE 3.5.0 beta series (was 3.4.1)

[Toni Wilen]

WinUAE 3.4.1 beta series

--

Main rule: Before reporting, make sure bug is NOT in previous offical version. It must be unique to this beta series! (If it is -> post in separate thread). No feature requests.

More information: http://eab.abime.net/showpost.php?p=861456&postcount=1.

--

http://www.winuae.net/files/b/winuae_3410b1.7z
http://www.winuae.net/files/b/winuae64_3410b1.7z

All kinds of misc changes and updates.

Beta 1:

- Quickstart Blizzard modes (except if JIT enabled) enabled 24-bit addressing which caused new autoconfig/fallback mode support to think Blizzard is in 68020 fallback mode, disabling autoconfig.
- Added Apollo 1240/1260 SCSI module installed on/off option.
- IDE Identify drive command didn't reset multiple sector offset value. If previous read was partial multiple transfer, identity drive returned previously read data block.
- Added Lock drive option to real harddrive mount dialog. If checked, all partitions in selected drive will be locked and unmounted when emulation is running, enabling full read/write access even if drive has Windows mounted FAT partition(s). Lock will be ignored if drive has one or more NTFS partitions.
- Add harddrive didn't accept magneto-optical devices.
- Lots of softfloat FPU mode and normal FPU mode fixes and updates. (Andreas Grabher and me)
- Softfloat FPU mode added to GUI, supports on the fly switching.
- Untrap middle mouse button option was stuck.
- Implemented accurate 68020+ T0 trace mode emulation. (Original one has not been changed for ages..)
- Fixed crash when sound card audio play started and mono sound channel mode and "Include CD and FMV audio" was ticked.
- 271b3 "Blitter final D write don't increase "nasty" count" That is not true, I think this was only added to fix demo that randomly glitches on real hardware too.. Fixes intro Himalaya by Avalanche.
- Added illegal instruction at the end of each memory bank allocation, just outside of last available byte to force exception if execution falls of the edge of memory. This can happen if "more compatible" is not checked. Currently this causes bus error exception. May become halt in the future.
- Audio interrupts are delayed by 2 cycles, logic analyzer confirmed. Emulated only in cycle-exact mode. (Mission Elevator)
- Emulated sprite special case when DMA mode sprite's start X-coordinate is less than sprite's DMA slot position.
- Fixed AGA HAM6/8 behavior when BPLCON4 XOR value is non-zero. Planes 6 and 7 never affect palette selection if HAM8, Planes 4 to 7 if HAM6. If HAM control bits select hold mode, selected direct R/G/B value bypasses XOR operation. (Thanks to dissident for test cases)
- AGA EHB plane 6 half-brite selection bit is taken before BPLCON4 XOR adjustment. (Same)
- HAM mode and BPLCON4 BPLAM bit changes or HAM mode and BPLCON4 sprite bank bit changes and sprites in same scanline may have caused wrong sprite or HAM colors.
- >2M chip RAM didn't work in non-JIT compatible modes.
- Added -ldlog 1/2 command line parameter, logs all laserdisc commands.
- Recursive mode ROM scanner now skips directories starting with dot.
- HDF creator used old style dialog type.
- 68030 CE/prefetch mode cache access fix.
- JIT direct broke 512k and 1M chip RAM size detection.
- If statefile is loaded with one or more floppy images that can't be opened: keep fake disk in drive (like previously) and ask for new disk path when missing disk is accessed for the first time. (access = read or write attempt) Old disk path is shown in disk dialog title bar.
- SCSI emulation SCSI buffer re-sizing didn't handle READ(6) zero size length correctly.
- Added memory cycle-exact Quickstart step for A1200 and CD32 configurations. Less CPU heavy than full cycle-exact and usually not much more worse (or better). At least not until 68020 CE gets better.
- Memory cycle exact mode incorrectly counted internal CPU cycles partially.
- Memory cycle exact mode also enables more accurate custom chipset mode.
- 68020 CE mode adjustments, fixed Chip RAM access speed (was too fast previously), removed all extra cycle counting temporarily. Work in progress.
- Include key scancode in Input panel key events.

Softfloat FPU mode is now officially supported, Andreas Grabher (Previous) implemented required 6888x-specific features to softfloat support library. Only needed if "weird" features are required (denormals and unnormals, invalid operands) or extended precision (that WinUAE does not support due to MSVC not supporting it) or accurate status register flags or 100% accurate rounding/NaN handling, FMOD/FREM quotient etc.

Arithmetic exceptions are not yet supported. Trigonometric and logarithmic functions are currently implemented using normal native C-library functions and won't return 100% same results as real 6888x.

Not compatible with JIT FPU emulation.

[Cylon]

- Include key scancode in Input panel key events.

Will try, Thank you very much, Toni!

[AMIGASYSTEM]

Thanks Toni for this new beta.

Untraps Mouse it was very convenient, ALT+TAB it conflicts with applications Windows minimized, when you press ALT+TAB It opens the program Windows parked on the taskbar.

[amilo3438]

Originally Posted by Toni Wilen

- Added memory cycle-exact Quickstart step for A1200 and CD32 configurations. Less CPU heavy than full cycle-exact and usually not much more worse (or better). At least not until 68020 CE gets better.
- Memory cycle exact mode incorrectly counted internal CPU cycles partially.
- Memory cycle exact mode also enables more accurate custom chipset mode.
- 68020 CE mode adjustments, fixed Chip RAM access speed (was too fast previously), removed all extra cycle counting temporarily. Work in progress.

Nice to see something is happening in this area!

I have checked some of remaining problematic A1200 demos in the quickstart basic config; seems that some of 50% demos works now better. Only the demo "C_LOUS-Cosmicglide" got some new graphics glitches that did not existed before.

[SaphirJD]

That's awesome, if demo compatibility can actually also a bit improved for more demanding configurations, looking forward for more cool stuff

[Toni Wilen]

Originally Posted by AMIGASYSTEM

Untraps Mouse it was very convenient, ALT+TAB it conflicts with applications Windows minimized, when you press ALT+TAB It opens the program Windows parked on the taskbar.

What do you mean? Nothing has changed. (EDIT: option works again, it was stuck on previously)

[AMIGASYSTEM]

Yes ALT+TAB It is not changed, i did not use for the cited reason,
was using the central button mouse which was very convenient.

[Toni Wilen]

Originally Posted by AMIGASYSTEM

was using the central button mouse which was very convenient.

Huh? Nothing has changed. Just switch it on!

[AMIGASYSTEM]

Originally Posted by Toni Wilen

- Untrap middle mouse button option was stuck.

Toni i mean this, i used middle mouse button, ever used Alt-Tab for the reasons mentioned

EDIT:
Toni sorry I found out that it was only disabled in "Miscellanous"

[1time]

Thanks Toni!

[White]

Hello Toni, thanks for the new 3.4.1 beta
OS4.1 freezes every now and then, do not crash
I can send a log if they can be useful? which?
the system does is copying files, that even in simple file information operations.
thank you

[emufan]

Originally Posted by AMIGASYSTEM

Toni i mean this, i used middle mouse button, ever used Alt-Tab for the reasons mentioned

EDIT:
Toni sorry I found out that it was only disabled in "Miscellanous"

of course something ahs changed, aas seen the other day before:
- p->input_mouse_untrap = MOUSEUNTRAP_MIDDLEBUTTON;
+ p->input_mouse_untrap = MOUSEUNTRAP_NONE;

[Toni Wilen]

Originally Posted by White

Hello Toni, thanks for the new 3.4.1 beta
OS4.1 freezes every now and then, do not crash
I can send a log if they can be useful? which?
the system does is copying files, that even in simple file information operations.
thank you

Unfortunately this is most likely normal. It can happen since uae expansion support was added, only way to make it happen less commonly is to have fast pc

Originally Posted by emufan

of course something ahs changed, aas seen the other day before:
- p->input_mouse_untrap = MOUSEUNTRAP_MIDDLEBUTTON;
+ p->input_mouse_untrap = MOUSEUNTRAP_NONE;

[/quote]

I forgot that default was also changed. Anyway, jumping to conclusions instead of trying the option first is not recommended. Or at least asking.

[mark_k]

Originally Posted by Toni Wilen

I forgot that default was also changed. Anyway, jumping to conclusions instead of trying the option first is not recommended. Or at least asking.

Would you be able to change the default back to enabled? Otherwise, for every config I'll have to enable middle-button-untraps and save it. [Before when the default was middle-button-untrap enabled, that line wasn't saved to config files. So when you load a config into 3.4.1b1 the option is disabled.]

[AMIGASYSTEM]

Originally Posted by mark_k

for every config I'll have to enable middle-button-untraps and save it.]

What I did on all my setups

[Toni Wilen]

Originally Posted by mark_k

Would you be able to change the default back to enabled? Otherwise, for every config I'll have to enable middle-button-untraps and save it. [Before when the default was middle-button-untrap enabled, that line wasn't saved to config files. So when you load a config into 3.4.1b1 the option is disabled.]

That was bad default and now it causes problems when other similar options are available. Unfortunately middle button option was saved as "optional" option which now causes another problem.

I am not sure how to fix this..

[Tomislav]

I made default.uae containing only 1 line:

win32.middle_mouse=true

[Toni Wilen]

Don't hack your config files too much, there will be solution.. (I hope)

[Toni Wilen]

http://www.winuae.net/files/b/winuae_3410b2.7z
http://www.winuae.net/files/b/winuae64_3410b2.7z

Beta 2:

- FM801 16-bit audio corruption fixed.
- Some 68020 CE mode cycle-counting is back.
- Clear button state mask when mouse capture state changes. If mouse or joystick button was kept pressed when mouse was uncaptured, button was kept pressed when capture state was restored even if real button was already pressed.
- G-REX and Cirrus Logic graphics board state was reset if autoconfig board info was re-scanned (entering RAM or HW Info panels) and emulation was already running (3.4.0)
- Middle mouse button default is back to enabled. (was changed in b1)
- Magic mouse on the fly on change didn't work.
- FPU exception handling rewrite, arithmetic exceptions partially implemented (Andreas Grabher), 68040 unimplemented FSAVE exception frame compatibility improved. Arithmetic exceptions are not that useful in Amiga emulation, programs that use FPU arithmetic exceptions are very rare. I don't know any, report if you find one
- FPU arithmetic exceptions checkbox added to GUI for easier testing. Will be removed when exceptions are fully implemented. Even when exceptions are disabled, they are still logged. Softfloat FPU only only.

[michaljarz]

Hi. 3.4.1b01 & b02 FPU with JIT is not working properly.
I used "Mand2000D" to generate fractals. Increase requiring fpu gave as a result a black screen (window).