Bit operation in fragment program?

OpenGL
1

Hi
I have to use 16 bits to my fragment program. Is there an elegant way to do that? It mean, I want to be then able to read each of this 16bits, easily (e.g: MUL myout, myin, bit[5])
Thanks

2

No shading languages have bitwise operations. Shader ops tend to be floating-point in nature.

3

Ok. But if I have 16 bits to access, I can put them in a float. And if I was in C++, I could simply use modulo operation to find the desired bit. Can I do the same in fragment program?
An other solution would be to pass the 16 bit in a texture, but a texture lookup is too slow and I can’t use it.
Thanks for your help

4

I seriously doubt it. The mod operator is an integer operation (technically not true, but close enough), and I don’t recall seeing any mention of such a function in the glslang spec. It certainly isn’t in any fragment program specs.

5

There is no mod operation in GLSL or in the asm languages, but integer operations are possible.
With a series of multiplication by 2 and division by 2, it’s possible to extract bits.

Unfortunatly, the overflow behavior is not portable and also, implementation are free to choose the integer precision they want.

6

Looking at the Orange book there is a mod() function. I was able to create a bit-wise testing function that returns the value of a desired bit. See the following:

float BitSelect(vec3 pixel, float bit)
{
  float value = 0.0;
  float bitlcl = bit;

  if (bit < 8.0)
    value = pixel.r;
  else if (bit < 16.0)
  {
    value = pixel.g;
    bitlcl = bit - 8.0;
  }
  else
  {
    value = pixel.b;
    bitlcl = bit - 16.0;
  }

  float ison = floor(mod(floor(value*255.0)/pow(2.0,bitlcl), 2.0));
  
  return ison;
}

Cheers,
-B

7

Hmmm. perhaps I was wrong back then.
Also, this thread is 1 year old. You will be able to do bitwise operations on the next generation of hw.

8

This topic was automatically closed 183 days after the last reply. New replies are no longer allowed.