Corrected shaders for WinUAE

[Dr.Venom]

Originally Posted by guest.r

It seems it's one of the nVidia/AMD differences again and thanks for reporting it. Is it fixed now for your setup?

Unfortunately not

I assume you upped the smartres version to exclude GammaOut for the time being? (I like the plain Hires version better, also because of GammaOut )

In attached zip saturation 1.0 and 2.0 are compared for the updated smartres version. As you can see still only part of the image is affected by the saturation, but not the reds. For reference I also included the full RA one with saturation at 2.0. Hope it helps..

[guest.r]

It's definitely something on the nVidia side. I rearanged the saturation calculations to do a seperate application per color channel. Hopefully it's fixed now.

[Dr.Venom]

Originally Posted by guest.r

It's definitely something on the nVidia side. I rearanged the saturation calculations to do a seperate application per color channel. Hopefully it's fixed now.

No change .

I also tried it without custom _WinUAE.fx and with D3D11, but makes no difference.

Would it be an idea to first isolate the saturation function in a separate saturation shader to see if the issue is not somewhere in the interaction with scanline function etc?

[Dr.Venom]

Originally Posted by guest.r

It's definitely something on the nVidia side. I rearanged the saturation calculations to do a seperate application per color channel. Hopefully it's fixed now.

Today I tried the same on an AMD (HD6850) setup, but it has exactly the same issue, see attached Saturation 1.0 versus 2.0 comparison. Only part of the image has saturation changed, red most noticably not changing.

I guess we can rule out AMD/NVidia issue. It seems either the shader saturation code is "broken", or possibly there's an issue with WinUAE shader compiler?

Could you make a version with saturation code completely removed? I would like to compare the output of that shader to the current one with saturation at 1.0. I have a hunch the output will differ...(?)

[guest.r]

Ah, i know the problem now. In the RA preset, saturation is applied after color-profiles. If it's applied before, the conversion matrices revert the red color in the 4-1 case. So i added the saturation parameter and function to the color-profiles too and it should work like with the crt-guest-dr-venom setup.

[Dr.Venom]

Originally Posted by guest.r

Ah, i know the problem now. In the RA preset, saturation is applied after color-profiles. If it's applied before, the conversion matrices revert the red color in the 4-1 case. So i added the saturation parameter and function to the color-profiles too and it should work like with the crt-guest-dr-venom setup.

Thanks for looking into it further, but that's not it.

I guess the naming convention from the files in the previous zip was confusing. I had included both a comparison for setting 1.0 and 2.0 without color profiles and with color profiles. If you look again at the two shots without color profiles from my previous post (they are named x_amd6850-SAT1.0.png and x_amd6850-SAT2.0.png) you'll see that the saturation for red does not change between setting 1.0 and 2.0 if you flip forward and backward between the two.

I would appreciate it if you could do the same exercise yourself, such that at least the full issue is clear.

It seems either the shader saturation code is "broken", or possibly there's an issue with WinUAE shader compiler?

Could you make a version with saturation code completely removed? I would like to compare the output of that shader to the current one with saturation at 1.0. I have a hunch the output will differ...(?)

[guest.r]

NP, saturation removed in one of the shaders. I just want to say that it puzzles me a bit, since it works normally here.

I'll add two screenshots of the color profile shader using the 4-1 configuration with saturation at 1.0 and 2.0.

[Dr.Venom]

Originally Posted by guest.r

NP, saturation removed in one of the shaders. I just want to say that it puzzles me a bit, since it works normally here.

Yeah, I can imagine. I got puzzled myself just now, as it seems my comments for the stuff without the color profiles is not right. I got confused myself with all the testing I guess (facepalm ).

Tonight I'll test the color-profiles stuff again, but I think the earlier solution you had worked out in the smart-res shader may have fixed it already, such that the addition of saturation to the color profiles shader itself may not even be needed... Anyway, I'll test it again tonight thoroughly just to be really sure. Thanks for bearing with me

[Dr.Venom]

Originally Posted by guest.r

Ah, i know the problem now. In the RA preset, saturation is applied after color-profiles. If it's applied before, the conversion matrices revert the red color in the 4-1 case. So i added the saturation parameter and function to the color-profiles too and it should work like with the crt-guest-dr-venom setup.

I guess you already knew, but this is indeed the proper fix . Works lovely now. I think the only "hassle" is that you need to change saturation in both the main shader and the color profiles shader to the same value if you want saturation a bit different.

Also happy to see that the color profile 4 works very well for Amiga / WinUAE in DCI-P3 mode. It's very nice to have the CRT and WinUAE run next to eachother and easily forget which is which. Even more so because the WinUAE shader and Amiga system and games were not part of the tests when creating the profiles, so it kind of works properly "out of sample", which is nice.

Could you update the shader pack with the fixed Color-Profiles.fx and the smartres shader that has GammaOut added? That way newer users will have everything handy with one zip. No scrolling through forums to get all the proper files

Speaking of which, I noticed the custom _WinUAE.fx is not part of the shader pack. Personally I think it's so valuable for proper looking CRT effect that it should be part of it? It would only need a very short readme, telling people to set the proper screen width in the shader, using it with D3D9 and set a dummy mask in the filter tab to make it work. Maybe add that too to make the pack complete?

Anyway, cheers again, back to playing some Amiga games!

[Dr.Venom]

Hi Toni,


I'm trying to switch shaders during runtime, using "displaydata" in the config file, as a tryout like the following:

displaydata=nlace,cmd=gfx_filter=D3D:CRT-Trinitron-SmartRes.fx
displaydata=lace,cmd=gfx_filter=D3D:CRT-Trinitron-HiRes.fx

Unfortunately this does not work, the screen goes black and from the log (attached) I see that it stays in a loop of applying the shader.

I tried this with both d3d9 and d3d11, same result (log is from d3d9)

Is this supposed to work (I guess maybe not..)?

If it's not supposed to work, is there an equivalent to use in your recent shader parser addition?

The goal is to switch shaders upon lace/non-lace switch, or alternatively a way to let the shader know that interlace is detected.

Would anything be possible with "displaydata=lace" in the "string winuae_config : WINUAE_CONFIG =".?

If so, could you possibly give a pointer what that should look like and what it would return to the shader?

Originally Posted by Toni Wilen

Above config in shader feature is now implemented.

Only first 15 lines are checked for "string winuae_config : WINUAE_CONFIG =".

[Dr.Venom]

@ guest.r

Have a question on the differences between the Trinitron versions .

Trinitron-HiRes vs. Trinitron-HiRes-SmartRes, what does the "smartres" do different for the hires version?

Trinitron-Hires-SmartRes vs. Trinitron-Smartres, which one would you recommend for games, or are they both equally suitable? I think they all look pretty good, just curious what the benefit "technically" is for the Trinitron-SmartRes version when playing games?

I'm a bit charmed by the hires-smartres version as I'm loading most games and demos through whdload, for which during boot (wb shell in hires) the screen look great also, opposed to the trinitron-smartres version.

In that regard, could you add the GammaOut to the Trinitron-SmartRes version? Then I can compare the apples to apples

Lastly, with Brightboost, what is the "neutral" setting for brightboost? Can it be compared to the brightness setting on a CRT where the "middle" is the default / neutral setting? If so, is the "1.0" setting for brightboost neutral?

[guest.r]

@Dr. Venom:

Trinitron-HiRes vs. Trinitron-HiRes-SmartRes, what does the "smartres" do different for the hires version?

It's a whole different horizontal/vertical algorithm, "hires" version only works better on interlaced screens and, in theory, on small details with productivity screens.

Trinitron-Hires-SmartRes vs. Trinitron-Smartres, which one would you recommend for games, or are they both equally suitable? I think they all look pretty good, just curious what the benefit "technically" is for the Trinitron-SmartRes version when playing games?

They are both suitable for games, OTOH you can get softer image with Trinitron-Smartres, and a bit of glow is added. Trinitron-Hires-SmartRes works well with games like Apidya, which is partially in hires. I like the benefit of having WB displayed correctly in hires mode also.

Lastly, with Brightboost, what is the "neutral" setting for brightboost? Can it be compared to the brightness setting on a CRT where the "middle" is the default / neutral setting? If so, is the "1.0" setting for brightboost neutral?

It's actually hard to tell the correct brightboost value due the way the scanlines are calculated. 1.0 is the starting point. Crt Hillyan goes as far as 1.5 by default. it's a bit different in gamma space where it's applied and higher values aren't as terrible as they look. Around 1.2 should be normal i think.

Last but not least, new version of the pack is here:
- custom _winuae.fx added with some description,
- crt-shifty(2) added,
- some small tweaks like gammaOUT...

[Toni Wilen]

Originally Posted by Dr.Venom

Hi Toni,
I'm trying to switch shaders during runtime, using "displaydata" in the config file, as a tryout like the following:

Looks like it sort of works if there is no filter set when it triggers. Even if it worked, it isn't that useful because it always puts the filter in first empty slot..

Better design is needed..

[mcbpete]

Hey guest.r - What resolution and size settings do you recommend for your shaders? I'm have WinUAE running in fullscreen at 1920x180 and on the filter tab: Scaling ='Fullscreen (TV)', Horiz Size & Vert size = 'FS', Aspect Ratio Correction = 'Automatic'

Unfortunately on most on your smartres shaders the scanlines (or the pixels themselves) always seem kinda uneven (e.g. using the otherwise fabulous trinitron, some vertical scanlines are thin, some 'fat') no matter what number I try in the two sliders to the right of the Horiz & Vert size dropdown.

A picture is probably better at explaining, it's properly zoomed in but if I scale it down the resize interpolation might break what I actually see on the screen!

[Retro-Nerd]

Select 1x/4x for the shader resize. Works perfectly fine. And no Scaling at all. Use 2x/2x zoom for Hires and 4x/4x for LoRes shader versions.

[guest.r]

As Retro-Nerd wrote...

Don't forget to use Bilinear from Point/Bilinear from the Extra Settings menu.

[Retro-Nerd]

Ah yes, this is very important. Without the bilinear option you never get it right.

[mcbpete]

Originally Posted by guest.r

As Retro-Nerd wrote...

Don't forget to use Bilinear from Point/Bilinear from the Extra Settings menu.

Got it, looks really good - thanks both ..... however my CPU is now feeling a little erm, bothered! (A leap from <10% at 1x * 1x to 379% at 2x * 4x according to the GUI ). Guess I might need a new GPU for such prettiness!

[Retro-Nerd]

A combination of beam racing/lagless vsync and shader is a bit heavy on CPU/GPU. Without beam racing it's OKish. 1x/4x shader resize should be also much better than 2x/4x, performance-wise.

[mcbpete]

There's a weird thing I have with my GPU and beam-racing (I believe DamienD has the same issue - and GPU!) - Despite *everything* set to PAL (and living in a PAL country), turning on V-Sync or beam racing makes WinUAE run at 60fps (though not solidly, it randomly flickers between around 55-65). Presumably it's due the laptop monitor being locked at 60Hz - unfortunately it's not changeable even when digging deep in GPU and monitor settings.