Cg to GLSL shader script regressions

Maister - making a log here of all the shaders that broke down on Nvidia Shield after you did your last few updates to the script -

4xSoft family - Broken geometry/viewport - pretty similar to this actually -

http://s24.postimg.org/rezupyt79/Screenshot_2013_08_16_14_48_58.png

5xBRv3.7c-lq - Same deal there.

These shaders previously ran fine until the most recent updates you pushed.

Seems like regression tests are in order here before we pile on even more stuff to the shader script - otherwise it’s shooting totally in the dark.

Should be fixed some days ago. Some shaders are using uniform half4x4 modelViewProj instead of float4x4. I have to manually insert some pragmas in Cg source to avoid transpose hack which broke powervr.

OK, I’ll re-run the shader conversion script again today and report back.

False positive indeed - shaders are fine after having reconverted them with the latest version of the script.

Could you upload a zip file with the working shaders?

I have no programming knowledge and don’t under how to convert CG shaders to GLSL

@blackie417 Here you go: http://www.mediafire.com/download/5pgrq … s_glsl.zip

I am still having the same issue unfortunately : (

The shader is off center for some reason on the Nvidia Shield

Is there a simple to use program that will convert CG to GLSL? I have the nvidia toolset but I don’t understand how to run the script with it.

As you’ve discovered, the conversion script isn’t perfect. I’m not sure if the issues with those shaders (generally HQx and xBR) is a problem with the script itself or cgc (the part of the Nvidia Cg Toolkit that does the heavy lifting here) spitting out garbage output. Unfortunately, it’s hard to tell whether the cgc output is broken or not because it always outputs crazy, unreadable spaghetti code, even when it works…

damn too bad. It would be awesome to get the HQx style shaders working for retro games.

Thanks for trying though, I appreciate it.

Is this HQ2x? It used to have issues with an old version of cg2glsl.py where it wouldn’t transpose the MVP matrix correctly, but that was fixed a long time ago.

Hmm. Those were converted using a fresh pull from git, but there’s always the possibility that I botched something.