GLSL equivalent of Cg bool4

dugan · 15 January 2017 03:50

I’m hand-converting some CG shaders to GLSL for a project I’m doing, and the XBR shaders have a construct that’s giving me a bit of trouble.

bool4 value = a < b;
if (!value)

I assume that the GLSL equivalent is one of the following, but I’m not sure which one it is, and I was hoping someone here could tell me.

Is it is?

bvec4 value = lessThan(a, b);
if (!any(value))

Or is it this?

bvec4 value = lessThan(a, b);
if (!all(value))

Thanks.

Hyllian · 15 January 2017 06:51

Could you paste the specific shader you’re trying to port and the lines where that cosntruction is?

Thats because It depends on the IF construction you are trying to port.

‘if (!value)’ isn’t the same as ‘x = (!value) ? y : z;’. So I can’t answer exactly what you’re asking.

dugan · 15 January 2017 06:51

It’s xbr-lv2.

One line (condensed a bit for this post) is:

bool4 interp_restriction_lv0 = // snip
bool4 edri = (wd1 <= wd2) && interp_restriction_lv0;

Another line is long, so I’ll summarize:

It calls eq, which is returns a 4-component boolean vector:

bool4 eq(float4 A, float4 B)
{
	return (df(A, B) < float4(XBR_EQ_THRESHOLD));
}

And then that line has tests AND’d together, such as:

if (!eq(f,b) && !eq(h,d) || eq(e,i) &&

‘if (!value)’ isn’t the same as ‘x = (!value) ? y : z;’.

Thanks. I didn’t know that.

Hyllian · 15 January 2017 06:52

Ok, I ported exactly this xbr-lv2 to glsl to be used by ePSXe. So I think you can grab the code from it.

I converted all bools and bvecs to float and vecs because glsl is an annoying language, almost non-intuitive. Cg is so straight forward… Shame that it is abandoned now.

Get it in attachment.

dugan · 15 January 2017 06:52

Wow, thanks. That helps a lot. Going through that, one thing I realized pretty quickly is that I’m nowhere near the level where I could have done that on my own.

Here are two attempts. The vertex shader is the same between each attempt:

https://github.com/duganchen/dosbox_shaders/blob/master/xbr-lv2.vert

The first attempt at the fragment shader is a straight adaptation of the epsxe shader, and it works well:

github.com

duganchen/dosbox_shaders/blob/60a764a93e5d23242bed04ea555b6d7855afc6ce/xbr-lv2.frag

#version 330 core

/*
   Hyllian's xBR-lv2 Shader

   Copyright (C) 2011-2015 Hyllian - [email protected]

   Permission is hereby granted, free of charge, to any person obtaining a copy
   of this software and associated documentation files (the "Software"), to deal
   in the Software without restriction, including without limitation the rights
   to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
   copies of the Software, and to permit persons to whom the Software is
   furnished to do so, subject to the following conditions:

   The above copyright notice and this permission notice shall be included in
   all copies or substantial portions of the Software.

   THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
   IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
   FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE

This file has been truncated. show original

The second is an attempt at porting the one in the common-shaders repo, using what I learned from the epsxe shader, and… it really doesn’t work (there’s no noticeable effect at all).

github.com

duganchen/dosbox_shaders/blob/df1bcbc18020cf5dff685b367d06e6b989baa924/xbr-lv2.frag

#version 330 core

const float XBR_SCALE = 3.0;
const float XBR_Y_WEIGHT = 48.0;
const float XBR_EQ_THRESHOLD = 15.0;
const float XBR_LV1_COEFFICIENT = 0.5;
const float XBR_LV2_COEFFICIENT = 2.0;

/*
   Hyllian's xBR-lv2 Shader

   Copyright (C) 2011-2015 Hyllian - [email protected]

   Permission is hereby granted, free of charge, to any person obtaining a copy
   of this software and associated documentation files (the "Software"), to deal
   in the Software without restriction, including without limitation the rights
   to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
   copies of the Software, and to permit persons to whom the Software is
   furnished to do so, subject to the following conditions:

This file has been truncated. show original

I’m curious as to where I went wrong in my second attempt. One thing I noticed from the epsxe shader is that when you replaced the “mul” commands, you didn’t simply swap the order of the operands like I’d usually seen recommended. Did I mess up by simply swapping the operand order when I should have done what you did, or did I also mess up elsewhere? Do mat4x3(A, B, C, D) and float4x3(A, B, C, D) not actually create the same matrix?

Hyllian · 15 January 2017 06:52

Maybe you should invert this:

(XBR_Y_WEIGHT*Y) * mat4x3(B, D, H, F);

by this?

mat4x3(B, D, H, F) * (XBR_Y_WEIGHT*Y) ;

Well, I really don’t like how glsl treat matrix multiplications. It’s completely non-intuitive.

hunterk · 15 January 2017 06:52

I made a thread a looong time ago about converting GLSL to Cg (back before we had ALL TEH SHADERS) that could be helpful going the other direction: http://libretro.com/forums/showthread.php?t=402

You indeed have to swap columns with rows as one language is column-major (Cg, I think?) and the other is row-major (GLSL, I think?)

Hyllian · 15 January 2017 06:52

It seems to me that even inverting it wouldn’t work. You have to transpose the color matrix I think. In Cg there’s a transpose function: http://http.developer.nvidia.com/Cg/transpose.html. Maybe it’s the same for glsl.