Setting gl_FragDepth causes lag?

Question

I designed a very basic set of depth shaders for rendering depth to my shadow map depth textures. This is the code for the depth fragment shader that I used :

#version 330 core

in vec3 FragPos;
uniform vec3 lightPos;
uniform float farPlane;

void main()
{
    float depth = length(FragPos - lightPos);
    depth /= farPlane;
    gl_FragDepth = depth;
}

This code isn't much, it simply calculates the distance between a fragment and the light source, normalizes the value by dividing it by the light's far plane distance, and sets gl_FragDepth as the result.

The code works without any errors. I was testing the renderer with just two objects in the scene and one point light source. Later, I pulled a big interior scene, and the FPS dropped from somewhere between 60-70, down to 30-40.

I tried doing some GPU profiling with Nvidia Nsights, and discovered that the glDrawElements for my shadow pass was spending 4 ms. I zeroed down the problem to the final line of code in the fragment shader written above, gl_FragDepth = depth.

What i found out was that if I removed the expression gl_FragDepth = depth, the FPS jumped to 70s, with the draw call taking just 1 ms. Note that, everything else was untouched.

How could setting the gl_FragDepth value, cause low performance?

Answer 1

Writing to gl_FragDepth will disable early fragment tests:

Therefore, an implementation is free to apply early fragment tests if the Fragment Shader being used does not do anything that would impact the results of those tests. So if a fragment shader writes to gl_FragDepth, thus changing the fragment's depth value, then early testing cannot take place, since the test must use the new computed value.