Data corruption when replacing uniform array with 1d texture in WebGL

问题

I am doing some GPGPU processing on a large 4D input array in WebGL2. Initially, I just flattened the input array and passed it in as a uniform array of ints, with a custom accessor function in GLSL to translate 4D coordinates into an array index, as follows:

const int SIZE = 5; // The largest dimension that works; if I can switch to textures, this will be passed in as a uniform value.
const int SIZE2 = SIZE*SIZE;
const int SIZE3 = SIZE*SIZE2;
const int SIZE4 = SIZE*SIZE3;
uniform int u_map[SIZE4];

int get_cell(vec4 m){
  ivec4 i = ivec4(mod(m,float(SIZE)));
  return u_map[i.x*SIZE3+i.y*SIZE2+i.z*SIZE+i.w];
}

On the JavaScript side, the flattened data is passed in as follows:

const map_loc = gl.getUniformLocation(program, "u_map");
gl.uniform1iv(map_loc, data);

Where data is an Int32Array (because that's what's required to map onto GLSL ints) containing 8-bit values.

That works, but it severely limits that size of inputs that I can work with. Going up to a dimension size of 6 results in using 1296 uniform slots, when only 1024 are available for this data and other control inputs.

So, I want to switch over to using a texture to hold larger quantities of data. So, I've updated the JS code to this:

const tex = gl.createTexture();
gl.activeTexture(gl.TEXTURE0);
gl.bindTexture(gl.TEXTURE_2D, tex);
gl.texParameteri(gl.TEXTURE_2D, gl.TEXTURE_MAG_FILTER, gl.NEAREST);
gl.texParameteri(gl.TEXTURE_2D, gl.TEXTURE_MIN_FILTER, gl.NEAREST);
gl.texImage2D(gl.TEXTURE_2D, 0, gl.R8, data.length, 1, 0, gl.RED, gl.UNSIGNED_BYTE, data);

where data has been repackaged as a Uint8Array, which should be used to provide single-channel r values in a texture sampler object in GLSL. The GLSL code is updated as follows:

uniform sampler2D u_map;

int get_cell(vec4 m){
  ivec4 i = ivec4(mod(m,float(SIZE)));
  float r = texelFetch(u_map, ivec2(i.x*SIZE3+i.y*SIZE2+i.z*SIZE+i.w, 0), 0).r;
  return int(r);
}

which should grab the single valid channel value out of the underlying buffer, just as when we were using an array.

However, after doing this substitution, I just get junk. The values returned from get_cell appear to all be either 1 or 0--and none of them even reliably correspond to the presence of actual 1 or 0 values in the original data buffer.

What am I doing wrong?

回答1:

If the internal format of the texture is gl.R8, then the values which are returned at the texture look up by texelFetch, are in range [0.0, 1.0].

You've to multiply the values by 255.0, if you want to bet the values in the range [0, 255]:

int get_cell(vec4 m){
    ivec4 i = ivec4(mod(m,float(SIZE)));
    float r = texelFetch(u_map, ivec2(i.x*SIZE3+i.y*SIZE2+i.z*SIZE+i.w, 0), 0).r;
    return int(r * 255.0);
}

Note, the formats like R8, RGB8 and RGBA8 are unsigned normalized fixed-point format. The color values are assumed to be floating-point values in the range [0.0, 1.0], where the smallest possible value 0 is converted to 0.0 and the largest possible value (2^8-1 == 255) is converted to 1.0.

回答2:

You used the R8 format which is a normalized floating point format with values from to 0.0 to 1.1. If you want an int format then consider using R32I texture internalformat, supply your texture data as format gl.RED_INTEGER, type gl.INT and use a Int32Array, change your sampler to an isampler2D and use int r = textureFetch(....).r to read the data.

If you want an integer result you'll also need to make an integer texture for results and attach that to a framebuffer to render integer results.

来源：https://stackoverflow.com/questions/55713667/data-corruption-when-replacing-uniform-array-with-1d-texture-in-webgl

标签

gpgpu

webgl2