2. Synchronization

Use the source from the previous tutorial 1.

1 invokation call delivers this result:

glUseProgram(program);

glDispatchCompute(10, 1, 1);

glUseProgram(0);

However, calling the Compute Shader multiple times without synchronizing between the calls, the result will be undefined:

glUseProgram(program);

for (unsigned int i = 0; i < 50; i++)

{

glDispatchCompute(10, 1, 1);

}

glUseProgram(0);

To fix this, we have to synchronize between each invokation call:

glUseProgram(program);

for (unsigned int i = 0; i < 50; i++)

{

glDispatchCompute(10, 1, 1);

glMemoryBarrier(GL_SHADER_STORAGE_BARRIER_BIT);

}

glUseProgram(0);

Read more about it here:

https://www.opengl.org/wiki/GLAPI/glMemoryBarrier

If (for example) the compute shader writes into buffer object and you want to use those buffer objects as sources for vertex attributes to render a triangle, you will have to use:

• GL_VERTEX_ATTRIB_ARRAY_BARRIER_BIT