Uniform Variables are always 0 in Compute Shader

OpenGL
1

Hi,

I have trouble setting the uniforms nx, ny, nz, bx, by, bz in the following Compute Shader. I wrote their values in a buffer to check and they always have the value 0…

#version 450 core

#extension GL_NV_shader_thread_shuffle : require
#extension GL_NV_shader_thread_group : require

layout (local_size_x = 1, local_size_y = 1, local_size_z = 1) in;

layout (std430, binding = 0) buffer Dataset 
{
    float values[];
} dataset;

struct MinMaxPair
{
    float min;
    float max;
} minMaxPair;

layout (std430, binding = 1) buffer MinMaxArray
{
    MinMaxPair pairs[];
} minMaxArray;

uniform uint nx;
uniform uint ny;
uniform uint bx;
uniform uint by;
uniform uint bz;

void main()
{
    uint shader_x = gl_GlobalInvocationID.x;
    uint shader_y = gl_GlobalInvocationID.y;
    uint shader_z = gl_GlobalInvocationID.z;

    uint blockFlatIndex = shader_x + shader_y*gl_NumWorkGroups.x + shader_z*gl_NumWorkGroups.x*gl_NumWorkGroups.y;
    uint blockFirstPointIndex = shader_x*(bx-1) + shader_y*(by-1)*nx + shader_z*(bz-1)*nx*ny;

    MinMaxPair minMaxPair2;

    minMaxPair2.min = dataset.values[blockFirstPointIndex];
    minMaxPair2.max = dataset.values[blockFirstPointIndex];

    uint x, y, z;
    uint index;

    for (z = 0; z < bz; z++)
    {
        for (y = 0; y < by; y++);
        {
            for (x = blockFirstPointIndex; x < (blockFirstPointIndex + bx); x++)
            {
                index = x + y*nx + z*nx*ny;

                if (dataset.values[index] < minMaxPair2.min)
                {
                    minMaxPair2.min = dataset.values[index];
                }

                if (dataset.values[index] > minMaxPair2.max)
                {
                    minMaxPair2.max = dataset.values[index];
                }
                
            }
        }
    }

    minMaxPair2.min = bx;
    minMaxPair2.max = by;

    minMaxArray.pairs[blockFlatIndex] = minMaxPair2;
}

I am using Qt-Wrapper classes to interface with the shader program. The following code shows all that should be necessary to set uniforms…

Create the QOpenGLShaderProgram object
m_computeShaderProgram = new QOpenGLShaderProgram();

Add shader source code, link and bind the program.

if (!m_computeShaderProgram->addShaderFromSourceFile(QOpenGLShader::Compute, "./shaders/parallelMarchingBlocks/minMaxPairs.comp"))
    {
        std::cout << "minMaxPairs.comp compilation error: " << m_computeShaderProgram->log().toStdString() << std::endl;
    }

    if (!m_computeShaderProgram->link())
    {
        std::cout << "m_computeShaderProgram linking error: " << m_computeShaderProgram->log().toStdString() << std::endl;
    }

    if (!m_computeShaderProgram->bind())
    {
        std::cout << "m_computeShaderProgram bind error: " << m_computeShaderProgram->log().toStdString() << std::endl;
    }
    m_computeShaderProgram->release();

Bind shader program, set the uniform values and start the compute shader…

if (!m_computeShaderProgram->bind())
    {
        qDebug() << "ERROR WHEN TRYING TO BIND COMPUTE SHADER PROGRAM";
    }

    m_computeShaderProgram->setUniformValue(m_computeShaderProgram->uniformLocation("nx"), nx);
    m_computeShaderProgram->setUniformValue(m_computeShaderProgram->uniformLocation("ny"), ny);
    m_computeShaderProgram->setUniformValue(m_computeShaderProgram->uniformLocation("bx"), bx);
    m_computeShaderProgram->setUniformValue(m_computeShaderProgram->uniformLocation("by"), by);
    m_computeShaderProgram->setUniformValue(m_computeShaderProgram->uniformLocation("bz"), bz);
    context->glDispatchCompute(numBlocksXDirection, numBlocksYDirection, numBlocksZDirection);

I have checked with the debugger that the values that I am passing to setUniformValue are actually the values that I want to pass. I have also checked that uniformLocation() does not return -1 for the uniforms.

2

This is an OpenGL forum, not a Qt forum. You should include a dump of the underlying GL calls being made. They’re probably not what you’re expecting.

Also, you didn’t provide enough info to even infer what GL calls might be made by the above, because the types of C++ types of nx, ny, bx, by, and bz aren’t provided.

One more thing I noticed here:

So even if the C++ types of those vars are GLuint, this wrapper may not do what you want.

I would Check for OpenGL Errors. You may find you have some GL_INVALID_OPERATION errors queued.

3

Thanks for you’r reply. I will check the errors later when I have time.

Also, you didn’t provide enough info to even infer what GL calls might be made by the above, because the types of C++ types of nx , ny , bx , by , and bz aren’t provided.

My mistake. All of the arguments are of type GLuint

4

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