Pedro Santos

In jit.op, we can use a matrix as the second input OR a numeric value, usually float.

Is this possible at all using [jit.gl.slab @file op.*.jxs]? Or does the 2nd input only accept matrices?

In my case, I'm particularly interested in op.absdiff.jxs, but I think it's the same for all math operators.

I've resolved the situation doing the following, but would like to clean it up, if possible.

In jit.op, we can use a matrix as the second input OR a numeric value, usually float.
Is this possible at all using [jit.gl.slab @file op.*.jxs]? Or does the 2nd input only accept matrices?

different-behaviour-in-jit-op-and-op-shader

It seems that the shader "op.absdiff.jxs" for example expects a 2D matrix as its right operator. So in this case, I think it is not possible to send a float value, but a 2D matrix, the same way you do(see shader source code below).

Or you might try modifying the shader code so that it works with a single float value.

uniform sampler2DRect tex0; <<< main texture <<<<<<<<<<<<

uniform sampler2DRect tex1; <<< this is your comparison texture <<<<<<<<<<<<

    vec4 v0 = texture2DRect(tex0, texcoord0);

    vec4 v1 = texture2DRect(tex1, texcoord1);

    gl_FragColor = (result*outscale)+outoffset;

It seems that the shader "op.absdiff.jxs" for example expects a 2D matrix  as its right operator. So in this case, I think it is not possible to send a float value, but a 2D matrix, the same way you do(see shader source code below). 

Or you might try modifying the shader code so that it works with a single float value.
Hope it helps

varying vec2 texcoord0;
varying vec2 texcoord1; 

uniform vec4 inscale;
uniform vec4 inoffset;
uniform vec4 in2scale;
uniform vec4 in2offset;
uniform vec4 outscale;
uniform vec4 outoffset;

uniform sampler2DRect tex0; <<< main texture  <<<<<<<<<<<<
uniform sampler2DRect tex1; <<< this is your comparison texture <<<<<<<<<<<<

void main( void )
{
	vec4 v0 = texture2DRect(tex0, texcoord0);
	vec4 v1 = texture2DRect(tex1, texcoord1);

	v0 = (v0*inscale)+inoffset;
	v1 = (v1*in2scale)+in2offset;
	vec4 result = abs(v0 - v1);
	gl_FragColor = (result*outscale)+outoffset;
}


Looking at the shader code it becomes evident that a second texture is needed, although it would be useful if the behaviour was the same as jit.op.

As an aside, I guess Jitter technology is at a crossroad right now.

For years, Jitter objects were developed with the CPU in mind. Nowadays,

the logical trend, for performance reasons, would be to rely on the GPU/OpenGL to do the majority of matrix processing instead of the CPU, where applicable.

In my opinion, right now, the problem is that the accessibility of these shaders isn't the same as traditional jitter objects, since they are "encapsulated" in the jit.gl.slab object. Wouldn't it be nice if we had a direct [jit.gl.op] that worked like the traditional [jit.op]? We could see the shader parameters (attributes) by clicking in its input or going to the Inspector instead of having to open the .jxs file and look around.

What I mean is that I would love to see a more profound integration of the GPU slab processing with the Max interface.

Thanks for your response.
Looking at the shader code it becomes evident that a second texture is needed, although it would be useful if the behaviour was the same as jit.op.

As an aside, I guess Jitter technology is at a crossroad right now.
For years, Jitter objects were developed with the CPU in mind. Nowadays,
the logical trend, for performance reasons, would be to rely on the GPU/OpenGL to do the majority of matrix processing instead of the CPU, where applicable.

What I mean is that I would love to see a more profound integration of the GPU slab processing with the Max interface.


I agree it would be nice to have a small shader compiler inside the application( very much in the way vvvv handles their hlsl implementation). So far most of my glsl prototyping is made with qtz or with the shader builder, which means extra steps before implementing it inside jitter. However, I think slabs are a very nice solution to allow multipass shading.

Different behaviour in jit.op and op shader?