WinUAE 3.2.2 beta series

[Toni Wilen]

Quick 1 week beta series, mainly bug fixes.

Main rule: Before reporting, make sure bug is NOT in previous offical version. (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_3220b1.7z
http://www.winuae.net/files/b/winuae64_3220b1.7z

Beta 1:

- 68020+ BFFFO instruction's result value must be calculated from full offset (was offset AND 31), even when effective address is data register.
- When loading config with joystick that is not available, switch to keyboard layout that was previously selected or none, not layout A. Set layout A in default.uae if you want old behavior.
- If IPL rises during interrupt exception processing, new interrupt exception should start immediately when previous one was about to execute first instruction. Previously one instruction was always executed before new instruction exception processing started.
- Per-monitor high DPI update still caused repeated GUI window closing/opening.
- Slirp code had undefined C (unsigned <= 0 comparisons, in WinUAE only), probably accidentally worked when using older compiler. Caused slirp thread to use 100% CPU time continuously.
- Keyboard names are now listed in Input panel device list. (Replaces "HID Keyboard Device" or similar generic name)
- AGA FMODE register was reset to original value if state was restored and then display parameters changed (for example window was resized) and program didn't update FMODE for example in copper list. (3.1.0b10)
- AGA subpixel bitplane delay emulation rewritten. (Fire and Ice CD32 config screen jitter, Banshee flickering vertical line)
- Fixed wrong free call in function that creates zipped debugging log files.
- Renamed log file zip file name. includes now 64-bit identifier and version number.
- Added Windows build number to log version string.
- Filter reserved USB HID usage IDs, not just vendor reserved IDs.
- Changed HID RawInput handling, never unregister raw input, use rawinput device change events to detect device insertions/removals instead of generic change support that also includes other types of media changes. This update also works around odd Windows 10 (10586 only?) rawinput side-effect where device handle can change after RegisterRawInputDevices() call. Raw input device change notification support is Vista or later only.
- Updated built-in HRTmon to 2.36
- Clip out of bounds RTG coordinates instead of rejecting graphics operation completely.
- SSE2 rounding mode should match FPU rounding mode. (3.2.0b16)

[AnnaWu]

Thanks for the new update.

[sodapop]

Thank you Toni ! (+1 for my request !)

[B14ck W01f]

Originally Posted by Toni Wilen

Quick 1 week beta series, mainly bug fixes.

So we can expect another stable release next week?

[Michael]

What is expected in turms of speed of ppc cpu emulations?

Is 604 code much faster then 603 cpu emulator code?

There is np easy way to benchmark things....
but from 'real life' Quake1 test I get the following odd results

Quake1 68k/060 320x240 - 167 FPS
Quake1 WOS/603 320x240 - 12 FPS
Quake1 OS4/604 320x240 - 37 FPS

I guess the 68k-PPC bridge emulation makes things a lot slower under WOS
however, the host cpu is only 40$ loaded here during emulation of 2 cpus

[Toni Wilen]

So we can expect another stable release next week?

Thats the plan.

Originally Posted by Michael

What is expected in turms of speed of ppc cpu emulations?

Is 604 code much faster then 603 cpu emulator code?

There is np easy way to benchmark things....
but from 'real life' Quake1 test I get the following odd results

Quake1 68k/060 320x240 - 167 FPS
Quake1 WOS/603 320x240 - 12 FPS
Quake1 OS4/604 320x240 - 37 FPS

I guess the 68k-PPC bridge emulation makes things a lot slower under WOS
however, the host cpu is only 40$ loaded here during emulation of 2 cpus

It is QEMU. I don't know. I only know that FPU is not that fast because it is using softfloat library, not host FPU.

If you want to benchmark CPU, use CPU only benchmark, not something that also uses display and sound. Then run some program that only tests display and so on to get any real useful results, to see if there are some real bottlenecks or not. Using as many as possible _simple_ benchmarks.

m68k/PPC thread syncronization can waste lots of time if both threads need to access IO addresses continuously. Only RAM can be thread-safe.

(Not really beta related, moving soon)

[rbm]

It´s not important, but the PPC "RUN/STOP" information is not shown in the windowed mode status bar if a x86 bridgeboard is selected together a PPC board.
Maybe that is so, but seems that the information is overwritten.

[Toni Wilen]

Originally Posted by rbm

It´s not important, but the PPC "RUN/STOP" information is not shown in the windowed mode status bar if a x86 bridgeboard is selected together a PPC board.
Maybe that is so, but seems that the information is overwritten.

Not really 3.2.2 related but fixed.

[DamienD]

Originally Posted by Toni Wilen

- AGA subpixel bitplane delay emulation rewritten. (Fire and Ice CD32 config screen jitter, Banshee flickering vertical line)

Love your work as always Toni

This also fixes an issues with flickering when decrunching tracks in Super Skidmarks

[Toni Wilen]

http://www.winuae.net/files/b/winuae_3220b2.7z
http://www.winuae.net/files/b/winuae64_3220b2.7z

Beta 2:

- SASI/SASI (CHS) options were not visible in GUI when listing SASI controllers.
- No MIDI out devices detected, 64-bit only.
- AVIOutput with "capture before filtering": width is now always divisible by 16 and height is divisible by two.
- Added config file only optional AVIOutput position and offset parameters: recording_width and recording_height, (resolution of video) and recording_x and recording_y, offset from top/left of Amiga display. If offsets are not set or negative, Amiga display is automatically centered. Only available in "capture before filtering" mode. GUI setting may be added later.
- Windowed mode bottom bar CPU state didn't show PPC state if x86 and PPC CPU expansion was active at the same time.
- Phoenix Board SCSI emulation added.

Phoenix Board SCSI:
- Autoconfig, 5380 SCSI chip.
- Very simple design, PIO SCSI, software handshake. (No "fake DMA").
- Boot ROM version J added, pbscsi.device 3.1. ("Phoenix Board SCSI J 29/7/92")

[FOL]

Originally Posted by Toni Wilen

http://www.winuae.net/files/b/winuae_3220b2.7z
http://www.winuae.net/files/b/winuae64_3220b2.7z

Beta 2:

- SASI/SASI (CHS) options were not visible in GUI when listing SASI controllers.
- No MIDI out devices detected, 64-bit only.
- AVIOutput with "capture before filtering": width is now always divisible by 16 and height is divisible by two.
- Added config file only optional AVIOutput position and offset parameters: recording_width and recording_height, (resolution of video) and recording_x and recording_y, offset from top/left of Amiga display. If offsets are not set or negative, Amiga display is automatically centered. Only available in "capture before filtering" mode. GUI setting may be added later.
- Windowed mode bottom bar CPU state didn't show PPC state if x86 and PPC CPU expansion was active at the same time.
- Phoenix Board SCSI emulation added.

Phoenix Board SCSI:
- Autoconfig, 5380 SCSI chip.
- Very simple design, PIO SCSI, software handshake. (No "fake DMA").
- Boot ROM version J added, pbscsi.device 3.1. ("Phoenix Board SCSI J 29/7/92")

Hi Toni,

Im having trouble extracting zips.
I keep getting unable to allocate enough storage. Even though I have 44GB free. I know its not my setup as I can extract other zips fine. I actually only use 7Zip on my system.

[McTrinsic]

AWESOME!

Great!

THANK YOU!!



Cheers,
McT

Originally Posted by Toni Wilen

[...]
- Phoenix Board SCSI emulation added.

Phoenix Board SCSI:
- Autoconfig, 5380 SCSI chip.
- Very simple design, PIO SCSI, software handshake. (No "fake DMA").
- Boot ROM version J added, pbscsi.device 3.1. ("Phoenix Board SCSI J 29/7/92")

[Arnie]

winuae_3220b2.7z exrtacts fine for me. It must be something on your system. Do you actually mean .zip files that (for me) open with Windows Explorer or just .7z?

.zip files extract O.K with 7zip for me also.

[Toni Wilen]

Originally Posted by McTrinsic

AWESOME!

Great!

THANK YOU!!



Cheers,
McT

Emulation of board is the easy part (even easier if SCSI chip is already emulated), the (very) difficult part is finding ROM images..

[FOL]

Originally Posted by Arnie

winuae_3220b2.7z exrtacts fine for me. It must be something on your system. Do you actually mean .zip files that (for me) open with Windows Explorer or just .7z?

.zip files extract O.K with 7zip for me also.

Zips are fine, 7zip seem to be an issue. Never had any problem before.

EDIT:- Just downloaded latest 7Zip, which is .7z and it extracted fine, so its not my system.

[Arnie]

Very strange. Have you tried downloading winuae_3220b2.7z again and retrying?

Have you tried using Windows Explorer to extract the archive?

[FOL]

Originally Posted by Arnie

Very strange. Have you tried downloading winuae_3220b2.7z again and retrying?

Have you tried using Windows Explorer to extract the archive?

Found issue. It appears its compressed with an option that doesn't exist in the last version of 7Zip. Updated to latest and now its extracting.
I had a simular situation a few months back, where I selected 7zip only feature for a zip and it no one else could extract it unless they had PC with 7Zip.

[dlfrsilver]

Thanks Toni

[AMIGASYSTEM]

Originally Posted by FOL

Found issue. It appears its compressed with an option that doesn't exist in the last version of 7Zip. Updated to latest and now its extracting.

Strange i use an old version ofWinRAR 4.2 and I never had any problem to unpack archives7Zip

[Toni Wilen]

http://www.winuae.net/files/b/winuae_3220b3.7z
http://www.winuae.net/files/b/winuae64_3220b3.7z

Beta 3: (Possibly also final)

- Debugger thought slow ram was IO region if ECS Agnus.
- CDTV-CR mode CD led didn't flash when reading data, only when playing audio.
- Fixed linestate array buffer underrun if mode was interlaced. Most common side-effect was blanked screen in interlaced modes, especially in fields or fields+ mode, if build was full optimized.