What are the criteria for deciding to use a static variable for optimization?

Question

Here is an example code why I started thinking about this

void renderScene(void)
{
    //Clear all pixels
    glClear(GL_COLOR_BUFFER_BIT);
    // Question : Declaring posAttrib as a static variable is good? or bad?
    GLint posAttrib = glGetAttribLocation(programID, "vtxPosition");


    glVertexAttribPointer(posAttrib, 3, GL_FLOAT, GL_FALSE, 0, ((void*)(0)));
    glEnableVertexAttribArray(posAttrib);
    glDrawArrays(GL_POINTS, 0, 3);

    //Double buffer
    glutSwapBuffers();
}

This is about OpenGl, and renderScene() will be called whenever my screen renders. So I think if I declare posAttrib as a static variable, it will reduce unnecessary calls.

But I learned using static variable causes many problems so dont use it. I considered what would be best. So I searched about it and read those posts

Declaring variables inside loops, good practice or bad practice?

Optimization for global and static variables

static variables in functions

This post said but the initialization is likely to take place at each loop iteration. and you probably should not worry about this kind of optimization. And I've read many posts advising against using static variables.

But in other opinions,using static variables is recommended for actions that incur high memory copy costs.

I'm not sure when to use static variables based on my level of CS knowledge.

Should I measure the performance difference between using and not using static variables to make a decision? But I expect this method would be difficult to use as I collaborate with more people and the project grows larger. I want to have a basis for deciding whether or not to use it on my own when writing code that accounts for a small part of the entire project.

So my question is: What are the criteria to use a static variable for optimization?

I'd say don't worry too much about performance unless you are actually measuring it being a problem. And if it is a problem, implement both ways and measure the difference. When it comes to performance, people often say `Measure, don't guess`. — dan1st, Apr 07 '23 at 07:37
Your example is not a very good one. For shaders specifically you either specify the attribute location in advance, or you grab it once and then cache it for the rest of the shader's lifetime. — Botje, Apr 07 '23 at 07:40
All optimization should be based on measurement and analysis. Any general advise like "use static variables" needs to be taken with more than a grain of salt. — Ulrich Eckhardt, Apr 07 '23 at 08:16
Static variables are not for optimization, but for things that are *required* to keep their value between function calls. In C++ we often chose to make those variables class members instead of function local statics. — BoP, Apr 07 '23 at 08:21

score 2 · Accepted Answer · answered Apr 07 '23 at 11:55

A static variable access might take one or two clock cycles less than a non-static variable access. (I say 'might' because with modern pipelined CPUs in most cases it will not make any difference whatsoever.)

Optimizations that save one or two clock cycles here and there are almost never worth making, and certainly never if in doing so you then have to worry about managing the variable, avoiding mistakenly overwriting it, etc.

The only viable reason for making a variable static as an optimization is for the purpose of optimizing not the accessing of the variable, but the initialization of that variable. In other words, if the variable is static, then you might be able to initialize it only once, whereas if the variable is not static, then you will have to initialize it each time you want to use it.

However, you can also achieve this with a member variable of an object that lives for as long as your program runs. So, what matters is how you use the variable, not whether the variable is static or not.

What are the criteria for deciding to use a static variable for optimization?

1 Answers1