WinUAE 3.6.1 beta series

[Toni Wilen]

3.6.1 beta series. This is planned to quick update, mainly bug fixes (and new features that can't cause too bad side-effects..) but same plan hasn't always worked previously so we'll see..

3.6.0 and 3.6.1 beta only. This time official 3.6.0 reports are also accepted here. Must mention if it is 3.6.1 only or also in 3.6.0! Anything older: this is wrong thread!

http://www.winuae.net/files/b/winuae_3610b1.7z
http://www.winuae.net/files/b/winuae64_3610b1.7z

Beta 1:

- Added Elsat Mega Ram HD controller emulation.
- A.L.F.3 HD controller emulation
- Rewritten expansion hardware interrupt handling to make it more compatible with PPC emulation.
- Implemented experimental RTG mode background thread for Amiga side VRAM to host texture color space conversion/copy. Option in RTG Board panel.
- Fixed crash at startup if DXGI 1.4 was highest available version (Windows 10 build 10240 only)
- KS 1.2 boot hack caused a crash in some manually configured UAE boot rom configurations without at least one directory filesystem/harddrive.
- OCS/ECS only BPLCON2 with "illegal" PF1 or PF2 value is now accurately emulated in dual playfield mode. (If BPLCON2 PFx>=5 and matching playfield is not transparent: playfield is drawn using COLOR0 with playfield priorities still working normally)
- D3D11: switching off on screen leds on the fly left static led texture on screen.
- Floppy sound type selection was not loaded correctly from config file (3.6)
- Floppy sound empty drive volume level was used if disk was inserted before emulation was started. (old)
- HDF filesystem loader now skips HUNK_DEBUG and HUNK_SYMBOL hunks instead of aborting filesystem initialization.
- Direct3D11 fullscreen, enter GUI, press save screenshot button: nothing happened.
- Direct3D11 window mouse drag resize now refreshes the display only after resizing ends.
- Do not allow UAE RTG init to succeed if it is enabled but not autoconfigured. (Few versions ago introduced new autoconfig handling broke previous "is this configured test?")
- Some 68000 instructions didn't use correct reversed memory access order in long -(An) addressing mode. (MOVE EA,-(an), MOVEM, ADDX, SUBX)
- Reading real 68EC040 MMU enable/page size TC register always returns zero, emulation now does the same. (Real 68EC060 not yet tested)
- 68040 SRP and URP registers are full 32-bit, 68060 masks out low 9 bits (which are ignored anyway when calculating descriptor table addresses)
- NCR5394 emulation improved (A.L.F.3 compatibility), support use of "Transfer information" command to manually read status and message bytes instead of "Initiator command complete sequence" that was designed for this purpose and all previously emulated 5394 based controllers have used.
- NCR5394 emulation didn't handle situation where data was written to FIFO before true DMA write was started. FIFO contents should be written before DMA transfer starts. (Blizzard SCSI Kit IV in some situations, possibly others)
- Added "Identity" checkbox to Add harddrive panel. If ticked, real ATA identity data is used in emulated ATA device instead of generic UAE generated and drive appears exactly as it does in real hardware (identical name, geometry etc). If checkbox is disabled: used HD controller/adapter does not support reading ATA Identity or drive is not ATA or ATAPI.
- Read Identity now also shows CHS geometry, LBA and LBA48 (if supported), parsed from identity data (if available)
- kickstart_ext_rom and cart config entries support Arcadia and ALG ROM images.
- joyport2 -config entry incorrectly used port 3 slot.

Elsat Mega Ram HD:

- Basic PIO IDE controller.
- elsat.device "ElsatHD 1.0 (09 Oct 1992)"

A.L.F.3:

- SCSI, NCR5394 based.
- Fake DMA, MOVEMs used in transfer loop.
- No official ROM dump available, tested using loadable driver ROM hack.

[DamienD]

Originally Posted by Toni Wilen

- joyport2 -config entry incorrectly used port 3 slot.

Toni, what can I say!!!

You never stop my friend, your new beta obviously fixes the v3.6.0 Dyna Blaster issue that I reported in my thread today.

...but could I please ask you to kindly put this fix into the stable v3.6.0 version?

Reason being, I know that your *next* version cycle lasts at least 6 months...

I know it's personal for me; but for anyone that now has my collection, better if people don't have to use a *beta* and then see / click warnings / extra stuff that are standard when using a *beta* version.

With an updated v3.6.0, I can point to this updated stable version in my thread without anything further required

[Amiga68k]

Hi Toni,

Is there some additional logging parameter (besides -log -logfile) I could use to explicitly look at sound frequency output vs. RTG frequency in more detail?

I still experience the same issue with 3.6.1 (and 3.6.0):

Windowed D3D9 and D3D11: All OK
Fullscreen D3D9: OK, Fullscreen D3D11: Not OK

Not OK = Sound output pitch too low and sound distortion (cracking)

When comparing log entries at the moment, I can only see similar values when comparing D3D9 vs. D3D11.

When going Fullscreen with D3D9:

29-811 [2440 000-000]: Device name: 'Intel(R) HD Graphics 4600' 140013000f12e3.6
30-313 [2440 000-000]: D3D9Ex: 00000056 00020800 E0020000 00000720 ALPHA DYNAMIC SCANLINE
30-314 [2440 000-000]: D3D9Ex: PS=3.0 VS=3.0 1024*768*0p FS VS=0 B=1I 32-bit 0 (8192x8192)
30-342 [2440 000-000]: D3D9Ex: pixelshader filter 'C:\DATA\WinUAE\plugins\_winuae.fx':-1 enabled
30-355 [2440 000-000]: D3D9Ex: 1024*768 main texture, depth 32
30-357 [2440 000-000]: Buffer size (1024*768) RTG
30-368 [2440 000-000]: NTSC mode V=59.2804Hz H=15590.7473Hz (227x262+1) IDX=-1 (<?>) D=0 RTG=1/1
30-369 [2440 000-000]: RTGFREQ: 262*59.2804 = 15531.4666 / 59.3 = 262

When going Fullscreen with D3D11:

41-626 [1197 000-000]: D3D11 init start. (1024*768) (1024*768) RTG=1 Depth=32.
41-627 [1197 000-000]: D3D11 found matching fullscreen mode. SLO=1 S=0. Default refresh rate.
41-628 [1197 000-000]: D3D11 FindClosestMatchingMode1() 148500000/2970000=50.00 SLO=1 W=1024 H=768
.
.
.
42-077 [1197 000-000]: Buffer size (1024*768) RTG
42-084 [1197 000-000]: NTSC mode V=59.2804Hz H=15590.7473Hz (227x262+1) IDX=-1 (<?>) D=0 RTG=1/1
42-084 [1197 000-000]: RTGFREQ: 262*59.2804 = 15531.4666 / 59.3 = 262
42-102 [1198 000-000]: D3D11 Shader and extra textures restored

I am wondering if additional logging in relation to output frequency might give a clue as to why this issue is occurring on my setup.

Currently it seems that only D3D11 'sees' that my Host Display panel is configured to output a 50Hz image (148500000/2970000=50.00) and does 'something'... (that's a pretty vague statement I'm afraid).

[ross]

Originally Posted by Toni Wilen

- Some 68000 instructions didn't use correct reversed memory access order in long -(An) addressing mode. (MOVE EA,-(an), MOVEM, ADDX, SUBX)

Hi Toni and thanks for the new beta.

Can you make an example of those affected instructions?
Are they really so rare that no software uses them ?

[Toni Wilen]

Originally Posted by DamienD

T...but could I please ask you to kindly put this fix into the stable v3.6.0 version?

Reason being, I know that your *next* version cycle lasts at least 6 months...

Not going to happen. But it is guaranteed to take much less than 6 months (check the first post again)

Originally Posted by Amiga68k

Currently it seems that only D3D11 'sees' that my Host Display panel is configured to output a 50Hz image (148500000/2970000=50.00) and does 'something'... (that's a pretty vague statement I'm afraid).

Does changing refresh rate in display panel change anything? (It is still quote odd that 50Hz is returned when asking for default. It should be highest supported normally)

Originally Posted by ross

Can you make an example of those affected instructions?
Are they really so rare that no software uses them ?

It means when writing long word, low word is written first, then high word. (It was low first, then high). Many -(an) address mode instruction use reverse write order on 68000.

No program notices any difference, unless it is some really tricky self-modifying code which most likely don't even exist. I guess at least in theory someone could do move.l ea,-(an) where an = $dff08c. (Similar CLR.L $dff088 case was fixed many many years ago and some programs really did it..)

[ross]

Originally Posted by Toni Wilen

It means when writing long word, low word is written first, then high word. (It was low first, then high). Many -(an) address mode instruction use reverse write order on 68000.

No program notices any difference, unless it is some really tricky self-modifying code which most likely don't even exist. I guess at least in theory someone could do move.l ea,-(an) where an = $dff08c. (Similar CLR.L $dff088 case was fixed many many years ago and some programs really did it..)

Ok, perfectly clear.
Many thanks!

[DamienD]

Originally Posted by Toni Wilen

Not going to happen. But it is guaranteed to take much less than 6 months (check the first post again)

Sure, you are the master Toni and obviously I respect whatever decisions you make without question

I'll just update my thread and people will need to download / click through *beta* related messages after extracting the latest file *only once* if they want to play Dyna Blaster without issue in my collection

[Toni Wilen]

Originally Posted by DamienD

Sure, you are the master Toni and obviously I respect whatever decisions you make without question

I'll just update my thread and people will need to download / click through *beta* related messages after extracting the latest file *only once* if they want to play Dyna Blaster without issue in my collection

Yeah but just fixing this one would annoy others that have found worse bugs...

You can work around it by using custom mapping (both gameports remap and input panel remap should work).

[Amiga68k]

Originally Posted by Toni Wilen

Does changing refresh rate in display panel change anything? (It is still quote odd that 50Hz is returned when asking for default. It should be highest supported normally)

Interesting, it does!

50Hz PAL:
Sound frequency OK, sound output better, still distorted though

59Hz:
Sound frequency and output OK! Strange side-effect, Fullscreen image is 'streched'.

I'll try and see what the logs tell in these situations.

[Toni Wilen]

Originally Posted by Amiga68k

Interesting, it does!

50Hz PAL:
Sound frequency OK, sound output better, still distorted though

59Hz:
Sound frequency and output OK! Strange side-effect, Fullscreen image is 'streched'.

I'll try and see what the logs tell in these situations.

Does http://www.winuae.net/files/b/winuae.7z select highest automatically when "default" is selected?

Refresh rate still shouldn't affect sound unless something causes vsync-like behavior.

[DamienD]

Originally Posted by Toni Wilen

Yeah but just fixing this one would annoy others that have found worse bugs...

You can work around it by using custom mapping (both gameports remap and input panel remap should work).

All good my friend; understand your reasons

Simple solution related to my collection if you want to play Dyna Blaster as you used to be able to before v3.6.0 was released:

http://eab.abime.net/showpost.php?p=...&postcount=320

[Amiga68k]

Originally Posted by Toni Wilen

Does http://www.winuae.net/files/b/winuae.7z select highest automatically when "default" is selected?

Refresh rate still shouldn't affect sound unless something causes vsync-like behavior.

Hah, getting there Toni!

With winuae64_3610b1:

40-087 [1170 000-000]: D3D11 init start. (1024*768) (1024*768) RTG=1 Depth=32.
40-088 [1170 000-000]: D3D11 found matching fullscreen mode. SLO=1 S=0. Default refresh rate.
40-089 [1170 000-000]: D3D11 FindClosestMatchingMode1() 148500000/2970000=50.00 SLO=1 W=1024 H=768
40-089 [1170 000-000]: D3D11 Device: Intel(R) HD Graphics 4600 [\\.\DISPLAY1] (0,0,1920,1080)

Log is also filled with the following:

43-955 [1328 000-000]: D3D11 Present 887a0001
.
.
.
43-955 [1475 000-000]: D3D11 Present 887a0001

(winuaelog_winuae64_3610b1.txt attached)

With your latest beta (renamed to winuae_3610b1):

03-783 [1135 000-000]: D3D11 init start. (1024*768) (1024*768) RTG=1 Depth=32.
03-785 [1135 000-000]: D3D11 found matching fullscreen mode. SLO=1 S=0. Default refresh rate.
03-785 [1135 000-000]: D3D11 Highest freq: 78750000/1049600=75.00 W=1024 H=768
03-787 [1135 000-000]: D3D11 FindClosestMatchingMode1() 78750000/1049600=75.03 SLO=1 W=1024 H=768

Sound frequency (pitch) and output in Fullscreen: OK.

(winuaelog_winuae_3610b1.txt attached)

Image in Fullscreen is still stretched though (it has correct Aspect-Ratio using D3D9...).

[bagpipes]

Wow, that was quick! Toni, in Arcadia cart config works perfect but what are the proper shortcuts for kickstart_ext_rom? For example: v3.00 ARCADIA doesn't work.

[AZka]

Thank You!

[Amiga68k]

Toni,

Symptom: Fullscreen D3D11 720x576 resolution has correct Aspect-Ratio, Fullscreen D3D11 1024x768 resolution image is stretched.

I did a comparison between the two. Log entries when going Fullscreen with 720x576:

39-668 [471 000-000]: D3D11 free start
39-668 [471 000-000]: D3D11 freed3d start
39-668 [471 000-000]: D3D11 freed3d end
39-674 [471 000-000]: D3D11 free end
39-677 [471 000-000]: WM_ACTIVATEAPP inactive 00DC0EE2
39-677 [471 000-000]: D3D11_resize -1 0 0 (0)
39-680 [471 000-000]: D3D11_resize 0 1 -1 (0)
39-719 [471 000-000]: D3D11 free start
39-721 [471 000-000]: D3D11 freed3d start
39-722 [471 000-000]: D3D11 freed3d end
39-724 [471 000-000]: D3D11 free end
39-728 [471 000-000]: SetCursorPos 360x288
39-730 [471 000-000]: WM_ACTIVATEAPP active 00950A78
39-731 [471 000-000]: D3D11_resize 1 1 -1 (0)
39-744 [471 000-000]: D3D11 init start. (720*576) (720*576) RTG=1 Depth=32.
39-751 [471 000-000]: D3D11 found matching fullscreen mode. SLO=1 S=0. Default refresh rate.
39-751 [471 000-000]: D3D11 Highest freq: 148351648/2475000=59.00 W=720 H=576
39-752 [471 000-000]: D3D11 FindClosestMatchingMode1() 148351648/2475000=59.94 SLO=1 W=720 H=576
39-753 [471 000-000]: D3D11 Device: Intel(R) HD Graphics 4600 [\\.\DISPLAY1] (0,0,1920,1080)
39-758 [471 000-000]: D3D11CreateDevice succeeded with level 11.1. Hardware accelerated.
39-764 [471 000-000]: D3D11 2 00000002 00000057
39-766 [471 000-000]: D3D11_resize 1 1 0 (0)
39-769 [471 000-000]: D3D11_resize 0 1 1 (0)
39-770 [471 000-000]: D3D11_resize -> fullscreen
40-095 [471 000-000]: D3D11 freed3d start
40-097 [471 000-000]: D3D11 freed3d end
40-110 [471 000-000]: D3D11 resize 720x576, 720x576 2 00000002 FS=1
40-147 [471 000-000]: D3D11 initd3d start
40-154 [471 000-000]: -> -360.000000 288.000000 360.000000 -288.000000 1.000000 1.000000
40-155 [471 000-000]: D3D11 initd3d end
40-171 [471 000-000]: D3D11 720x576 main texture allocated
40-175 [471 000-000]: D3D11 Shader and extra textures restored
40-176 [471 000-000]: POS (0 0 720 576) - (0 0 720 576)[720,576] (0 0) S=720*576 B=720*576
40-177 [471 000-000]: 0 0 1 1
40-179 [471 000-000]: -> -360.000000 288.000000 360.000000 -288.000000 1.000000 1.000000
40-184 [471 000-000]: D3D11 init end
40-189 [471 000-000]: D3D11 720x576 main texture allocated
40-189 [471 000-000]: POS (0 0 720 576) - (0 0 720 576)[720,576] (0 0) S=720*576 B=720*576
40-191 [471 000-000]: 0 0 1 1
40-193 [471 000-000]: -> -360.000000 288.000000 360.000000 -288.000000 1.000000 1.000000
40-196 [471 000-000]: Buffer size (720*576) RTG
40-201 [471 000-000]: NTSC mode V=59.2804Hz H=15590.7473Hz (227x262+1) IDX=-1 (<?>) D=0 RTG=1/1
40-202 [471 000-000]: RTGFREQ: 262*59.2804 = 15531.4666 / 59.3 = 262
40-223 [472 000-000]: D3D11 Shader and extra textures restored

Log entries when going Fullscreen with 1024x768:

25-888 [1144 000-000]: D3D11 free start
25-889 [1144 000-000]: D3D11 freed3d start
25-889 [1144 000-000]: D3D11 freed3d end
25-895 [1144 000-000]: D3D11 free end
25-899 [1144 000-000]: WM_ACTIVATEAPP inactive 00570956
25-903 [1144 000-000]: D3D11_resize -1 0 0 (0)
25-904 [1144 000-000]: D3D11_resize 0 1 -1 (0)
25-946 [1144 000-000]: D3D11 free start
25-948 [1144 000-000]: D3D11 freed3d start
25-949 [1144 000-000]: D3D11 freed3d end
25-949 [1144 000-000]: D3D11 free end
25-951 [1144 000-000]: SetCursorPos 512x384
25-953 [1144 000-000]: WM_ACTIVATEAPP active 006109E0
25-955 [1144 000-000]: D3D11_resize 1 1 -1 (0)
25-970 [1144 000-000]: D3D11 init start. (1024*768) (1024*768) RTG=1 Depth=32.
25-972 [1144 000-000]: D3D11 found matching fullscreen mode. SLO=1 S=0. Default refresh rate.
25-972 [1144 000-000]: D3D11 Highest freq: 78750000/1049600=75.00 W=1024 H=768
25-973 [1144 000-000]: D3D11 FindClosestMatchingMode1() 78750000/1049600=75.03 SLO=1 W=1024 H=768
25-973 [1144 000-000]: D3D11 Device: Intel(R) HD Graphics 4600 [\\.\DISPLAY1] (0,0,1920,1080)
25-980 [1144 000-000]: D3D11CreateDevice succeeded with level 11.1. Hardware accelerated.
25-986 [1144 000-000]: D3D11 2 00000002 00000057
25-987 [1144 000-000]: D3D11_resize 1 1 0 (0)
25-988 [1144 000-000]: D3D11_resize 0 1 1 (0)
25-989 [1144 000-000]: D3D11_resize -> fullscreen
26-294 [1144 000-000]: D3D11 freed3d start
26-295 [1144 000-000]: D3D11 freed3d end
26-298 [1144 000-000]: D3D11 resize 1024x768, 1024x768 2 00000002 FS=1
26-319 [1144 000-000]: D3D11 initd3d start
26-333 [1144 000-000]: -> -512.000000 384.000000 512.000000 -384.000000 1.000000 1.000000
26-335 [1144 000-000]: D3D11 initd3d end
26-363 [1144 000-000]: D3D11 1024x768 main texture allocated
26-364 [1144 000-000]: POS (0 0 1024 768) - (0 0 1024 768)[1024,768] (0 0) S=1024*768 B=1024*768
26-365 [1144 000-000]: 0 0 1 1
26-368 [1144 000-000]: -> -512.000000 384.000000 512.000000 -384.000000 1.000000 1.000000
26-369 [1144 000-000]: D3D11 init end
26-378 [1144 000-000]: D3D11 1024x768 main texture allocated
26-378 [1144 000-000]: POS (0 0 1024 768) - (0 0 1024 768)[1024,768] (0 0) S=1024*768 B=1024*768
26-380 [1144 000-000]: 0 0 1 1
26-382 [1144 000-000]: -> -512.000000 384.000000 512.000000 -384.000000 1.000000 1.000000
26-382 [1144 000-000]: Buffer size (1024*768) RTG
26-389 [1144 000-000]: NTSC mode V=59.2804Hz H=15590.7473Hz (227x262+1) IDX=-1 (<?>) D=0 RTG=1/1
26-390 [1144 000-000]: RTGFREQ: 262*59.2804 = 15531.4666 / 59.3 = 262
26-409 [1145 000-000]: D3D11 Shader and extra textures restored

The only difference that I spot is with "D3D11 Highest freq":

720x576: 148351648/2475000=59.00
1024x768: 78750000/1049600=75.00

Going Fullscreen with D3D9 gives minimal output (Aspect Ratio is correct though):

57-693 [1160 000-000]: SetCursorPos 512x384
57-711 [1160 000-000]: Device name: 'Intel(R) HD Graphics 4600' 140013000f12e3.6
58-183 [1160 000-000]: D3D9Ex: 00000056 00020800 E0020000 00000720 ALPHA DYNAMIC SCANLINE
58-185 [1160 000-000]: D3D9Ex: PS=3.0 VS=3.0 1024*768*0p FS VS=0 B=1I 32-bit 0 (8192x8192)
58-212 [1160 000-000]: D3D9Ex: pixelshader filter 'C:\DATA\WinUAE\plugins\_winuae.fx':-1 enabled
58-221 [1160 000-000]: D3D9Ex: 1024*768 main texture, depth 32
58-225 [1160 000-000]: Buffer size (1024*768) RTG
58-233 [1160 000-000]: NTSC mode V=59.2804Hz H=15590.7473Hz (227x262+1) IDX=-1 (<?>) D=0 RTG=1/1
58-236 [1160 000-000]: RTGFREQ: 262*59.2804 = 15531.4666 / 59.3 = 262

Could this very well just be a display driver bug?

[Toni Wilen]

Originally Posted by bagpipes

Wow, that was quick! Toni, in Arcadia cart config works perfect but what are the proper shortcuts for kickstart_ext_rom? For example: v3.00 ARCADIA doesn't work.

v3.00 ARCADIA works fine here. You sure you don't have some other related config entry that overrides it?

Originally Posted by Amiga68k

Toni,

Symptom: Fullscreen D3D11 720x576 resolution has correct Aspect-Ratio, Fullscreen D3D11 1024x768 resolution image is stretched.

Does display look the same if you select 60Hz from display panel refresh list? D3D11 (DXGI) can work differently than D3D9/DirectDraw because DXGI fullscreen requires refresh rate, it does not have similar concept of "default" that older APIs had.

Also visible size can change depending on refresh rate if display is not using digital connection and/or display is not fully digital. Or GPU control panel may have some weird scaling settings too.

[Amiga68k]

Originally Posted by Toni Wilen

Does display look the same if you select 60Hz from display panel refresh list? D3D11 (DXGI) can work differently than D3D9/DirectDraw because DXGI fullscreen requires refresh rate, it does not have similar concept of "default" that older APIs had.

Same result with D3D11:
Fullscreen 720x576 resolution has correct Aspect-Ratio
Fullscreen 1024x768 resolution image is stretched

Also visible size can change depending on refresh rate if display is not using digital connection and/or display is not fully digital. Or GPU control panel may have some weird scaling settings too.

I had to create the 720x576 resolution manually as a Custom Resolution within the Intel HD Graphics Control Panel (1024x768 already existed).

There is an overall setting within the Control Panel that has its setting on: 'Maintain Aspect Ratio'.

[Amiga68k]

EDIT:
Hmm, now getting some mixed results with 720x576 (I have Aspect ratio 4:3 in RTG board settings).

With 'official' 3.6.0 and 3.6.1 Beta 1:
Fullscreen has correct Aspect-Ratio

With the beta you provided the other day (post #10):
Fullscreen image stretched

1024x768 still stretched with all versions... argh!

EDIT2: Oops, this was supposed to be an addition to the message above

[Toni Wilen]

Include complete log file and config from all configurations (working and not working).

[Amiga68k]

Originally Posted by Toni Wilen

Include complete log file and config from all configurations (working and not working).

Please find first set of the 720x576 Fullscreen configuration attached.

3.6.0
Aspect ratio correct

3.6.1 Beta 1
Aspect ratio correct

3.6.1 Beta 1 from post #10
Image stretched