Is it worth caching glsl uniform location in code?

Question

I would like to have ability to set uniforms via their actual names in the shader

myProgram.uniform3fv("uniformVector", 0.0f, 0.1f, 1.0f);

do I have to cache locations in some form of a map?

std::map<std::string, unsigned int> // or unordered_map 

or maybe OpenGL (on desktop) caches this anyway, so I would not have any performance difference?

Colorful

Answer 1

The OpenGL specification defines functionality, not performance. So there's no way to know how any particular OpenGL implementation will store the list of active uniforms or how fast glGetUniformLocation will be, relative to std::map performance.

So really, it's up to you. If you want consistent, known performance, do it yourself. If you want to take a chance on the vagaries of the OpenGL implementation, then query it whenever you want. Personally, I'd say ditch names and go with locations. That way you get maximum performance.

Scrutinize

so as in most cases, we cannot rely on assumptions (about the driver implementation) and it is better to do such simple caching on our own. – Colorful