WinUAE 3.6.0 beta series

[Genju]

I've been having problems with some DX9 shaders in DX11 mode as well, xBR shaders especially. There are some fuzzy edges in the image (best seen on the Kickstart 1.3 startup screen) that aren't there in DX9 mode.

I've attached my test config, the shader file and winuaelog.txt.

[Toni Wilen]

Does it become correct if you do following change:

MinFilter = POINT;
MagFilter = POINT;

->

Filter = MIN_MAG_MIP_POINT;

[Genju]

Yes, seems fine now after making that change, thanks!

[Toni Wilen]

Perfect. I'll try to make this conversion automatic. (MinFilter/MagFilter=POINT/LINEAR to FILTER = MIN_MAG_MIP_POINT/MIN_MAG_MIP_LINEAR)

EDIT: Done.

[Toni Wilen]

Originally Posted by mark_k

Did you experiment with DXGI_SWAP_EFFECT_FLIP_DISCARD too?

Testing winuae.exe 2017-12-23 21:11 on Windows 10 1709, AMD FirePro W5100 in D3D11 mode:
- With "No buffering" vsync is now forced in full-window mode. Before (with 3.6.0b14) there was tearing in full-window mode with No buffering.
- With "Double buffering" there is still tearing in full-window mode (same as 3.6.0b14).
- Still in full-window mode, running emulation with double buffering, opening setttings, changing to no buffering and continuing emulation results in the previous image remaining static on-screen but sound playing. So emulation runs but the display is not updated at all.

I can't see any difference with DXGI_SWAP_EFFECT_FLIP_DISCARD (I was originally going to use it but RTG did partial updates and could not have worked, since then a lot has changed and now there is separate "staging texture" that custom emulation code updates, after update it is blitted to another texture that is used for rendering. Which mean _DISCARD will work fine in any mode)

Blit model is now only used if DXGI 1.5 tearing mode is not available.

[Toni Wilen]

http://www.winuae.net/files/b/winuae_3600b15.7z
http://www.winuae.net/files/b/winuae64_3600b15.7z

I think 3.6 is getting complete enough, estimated release date: January 2018.

Beta 15:

- Debugger 68040/060 MMU table dump command (mmud) didn't support 8k page size. Mask out page descriptor bit 3 (unused) and 4 (M) when checking for descriptor flag differences to generate more useful output from Fusion Forty boot ROM generated page tables.
- Debugger memory string routines now also use MMU mapping mode if enabled.
- Fusion Forty memory flag fixed (It does not use processor slot ram and it must not be enabled). Fixes boot crash if MMU emulation is enabled. Note that it will crash (nothing to do with emulation) if MMU is enabled and 32-bit RAM (except board's onboard RAM is enabled) or any Z3 boards are enabled. Real A2000 can't have those.
- D3D11 no-buffer mode now selects blit model (instead of flip model) if DXGI 1.5+ allow tearing mode is not available.
- D3D11 low latency vsync partially implemented. No calibration, it uses Windows reported value because they should be accurate (reported as fractional number)
- D3D11 black frame insertion support implemented. (which probably isn't working very well..)
- D3D11 mode will now reload shader(s) automatically on the fly if they are modified.
- D3D11 mode D3D9 to D3D11 compatible shader conversion now also converts D3D9-only migfilter/magfilter lines.
- D3D11 mode D3D9 to D3D11 compatible shader modification is completely skipped if technique10 or technique11 is found.
- D3D11 mode fullscreen alt-tab reliability improved, it was still possible to get blank or partially blank window after alt-tab.

[mark_k]

Originally Posted by Toni Wilen

Blit model is now only used if DXGI 1.5 tearing mode is not available.

Blit model allowing tearing (which I guess you see with Nvidia GPU?) doesn't seem to be universal. It may well be driver/hardware-dependent. Flip may always happen in vsync with AMD drivers. So on some hardware there will be no benefit to using blit vs flip. I wonder what the Intel graphics driver does.

[Saghalie]

Originally Posted by Toni Wilen

http://www.winuae.net/files/b/winuae_3600b15.7z
http://www.winuae.net/files/b/winuae64_3600b15.7z

I am getting a virus hit on winuae_3600b15.7z Type: Trojan:VBS/Foretype.A!ml

See attached.

[Toni Wilen]

Looks like weird Windows defender false positive because it reports .7z file as being a virus, not the contents of archive!

I disabled real-time scanning, downloaded it, scanned it manually: virus found, test 7z file, all ok, unpack 7z file, manually scan winuae.exe: nothing found.

virustotal.com does not report anything.

EDIT: I reported it using manual sample submit option.

[jotd]

running another OS than Windows on Windows is like a virus to Microsoft

[mark_k]

Originally Posted by Toni Wilen

- D3D11 low latency vsync partially implemented. No calibration, it uses Windows reported value because they should be accurate (reported as fractional number)

With low-latency vsync and "No buffering", emulation seems to run at double the correct speed. Happens in windowed, full-window and full-screen modes. Double and triple buffering seem fine. (AMD graphics card, Windows 10 1709.)

Edited to add...

With emulation running in D3D11 full-screen mode, press F12. Desktop reappears and settings window opens. Minimise the settings window. It moves to a small titlebar-height window at the lower left of the screen, just above the taskbar. (It doesn't minimise to the taskbar as it should.)

With emulation running in D3D11 full-screen mode, press F12. Desktop reappears and settings window opens. The settings window is always on top of other windows, even if "GUI window always on top" is not enabled in misc settings.

With emulation running in D3D11 full-screen mode, press F12. Click Reset. WinUAE returns to full-screen mode and emulation continues (the emulated machine resets of course), but the screen is just black. Press F12 again, see settings window on desktop. This time just click OK and emulation continues correctly.

[turrican3]

thank you tony for the last update,
i'll have a test.
sorry i'm too late for the logs... I'm a little busy those times.hopefully Genju was there.
edited: fast test, black screen with dx11, not picture at all but sound. link the logs beta 15
edited 2 : exactly same test but with beta 14, same result than before : dx11 display but badly. logs added. Perhaps it will be be easier to debug.

[Saghalie]

Originally Posted by Toni Wilen

Looks like weird Windows defender false positive because it reports .7z file as being a virus, not the contents of archive!

I disabled real-time scanning, downloaded it, scanned it manually: virus found, test 7z file, all ok, unpack 7z file, manually scan winuae.exe: nothing found.

virustotal.com does not report anything.

EDIT: I reported it using manual sample submit option.

Thanks!

[Toni Wilen]

Originally Posted by turrican3

thank you tony for the last update,
i'll have a test.
sorry i'm too late for the logs... I'm a little busy those times.hopefully Genju was there.
edited: fast test, black screen with dx11, not picture at all but sound. link the logs beta 15
edited 2 : exactly same test but with beta 14, same result than before : dx11 display but badly. logs added. Perhaps it will be be easier to debug.

shader.code(49,2-19): Deprecated DX9 state 'MinFilter' is being ignored for fx_5_0; use state 'FILTER' instead

Automatic conversion can't handle include files (it would not be good idea to modify them on disk automatically), you need to manually update shader.code:

MinFilter = LINEAR
MagFilter = LINEAR

->

Filter = MIN_MAG_MIP_LINEAR

and

MinFilter = POINT
MagFilter = POINT

->

Filter = MIN_MAG_MIP_POINT

If this does not help: attach config file.

[Toni Wilen]

Originally Posted by Toni Wilen

Looks like weird Windows defender false positive because it reports .7z file as being a virus, not the contents of archive!

I got mostly useless status report back which practically only says "No malware detected". I guess it was also fixed, 7z is now detected as clean.

[turrican3]

Originally Posted by Toni Wilen

Automatic conversion can't handle include files (it would not be good idea to modify them on disk automatically), you need to manually update shader.code:

MinFilter = LINEAR
MagFilter = LINEAR

->

Filter = MIN_MAG_MIP_LINEAR

and

MinFilter = POINT
MagFilter = POINT

->

Filter = MIN_MAG_MIP_POINT

If this does not help: attach config file.

config file attached + a dump file.
ps: i detected others problems. will report 2 night.
I hope you have everything you need.

[jotd]

Toni I think you fixed AGA display, cos a lot of AGA games had a trashed display with previous beta. Now it runs fine !

Also note that you fixed the crash on turning off bluetooth/inserting USB peripheral that I reported some time ago. Latest beta rocks!

thanks again for all this

[mcbpete]

In the latest beta build I've finally switched all my profiles from dx9 to dx11 and swapped the video filter from 'PAL' to 'd3d: Lanczos16' (seemed to 'blend' the colours and soften the pixels closer to how I remember things being on my TV back in the 90s!), and the CPU has dropped from ~40-50% right down to ~0-5% (frequently remains at 0%). Flipping magical witchcraft !

[jotd]

Holidays are always beneficial for amiga projects

[Amiga68k]

I have two configurations that I use on a regular basis:

'Games' setup: RTG 720 x 576
'Productivity' setup: RTG 1024 x 768

When I change Display settings of the first one to Full-window the mouse is 'off'. When clicking on the screen it 'activates' the click further to the right. Second symptom is I can move the mouse on the black parts of the screen on the left and right side.

With 1024 x 768 in Full-window mode, the mouse is correct and I cannot move the mouse to the left and right black parts.

Attached you can find the files that correspond to the first scenario.