Code 1:

#include <iostream>
#include <vector>

class Vector2d {
    public:
        double x, y;
        Vector2d(double x, double y) : x(x), y(y) { }
};

int main() {
    const double FoV = 50.0;
    std::vector<Vector2d> vecOfVector2d;
    vecOfVector2d.emplace_back(Vector2d(2.0, 3.0));
    vecOfVector2d.emplace_back(Vector2d(4.0, 5.0));
    
    for (std::vector<Vector2d>::const_iterator i = vecOfVector2d.begin(); i != vecOfVector2d.end(); ++i) {
        const Vector2d vec = *i;
        double xpos = rand() % 100; // Calculated but varies in each iteration
        
        auto isBBoxInsideFOVHorizontal = [FoV, xpos, vec]()->bool{
                std::cout << xpos << " " << vec.x << " " << vec.y << std::endl;
                return true;
        };
        
        isBBoxInsideFOVHorizontal();

    }
    
    return 0;
}

Code 2:

#include <iostream>
#include <vector>

class Vector2d {
    public:
        double x, y;
        Vector2d(double x, double y) : x(x), y(y) { }
};

int main() {
    const double FoV = 50.0;
    std::vector<Vector2d> vecOfVector2d;
    vecOfVector2d.emplace_back(Vector2d(2.0, 3.0));
    vecOfVector2d.emplace_back(Vector2d(4.0, 5.0));
    
    auto isBBoxInsideFOVHorizontal = [FoV](const double xpos, const Vector2d vec)->bool{ 
                std::cout << xpos << " " << vec.x << " " << vec.y << std::endl;
                return true;
    };
        
    for (std::vector<Vector2d>::const_iterator i = vecOfVector2d.begin(); i != vecOfVector2d.end(); ++i) {
        const Vector2d vec = *i;
        double xpos = rand() % 100; // Calculated but varies in each iteration
        
        isBBoxInsideFOVHorizontal(xpos, vec);
    }
    
    return 0;
}

Code 3:

#include <iostream>
#include <vector>

class Vector2d {
    public:
        double x, y;
        Vector2d(double x, double y) : x(x), y(y) { }
};

int main() {
    const double FoV = 50.0;
    std::vector<Vector2d> vecOfVector2d;
    vecOfVector2d.emplace_back(Vector2d(2.0, 3.0));
    vecOfVector2d.emplace_back(Vector2d(4.0, 5.0));
    
    for (std::vector<Vector2d>::const_iterator i = vecOfVector2d.begin(); i != vecOfVector2d.end(); ++i) {
        const Vector2d vec = *i;
        double xpos = rand() % 100; // Calculated but varies in each iteration
        
        auto isBBoxInsideFOVHorizontal = [FoV](const double xpos, const Vector2d vec)->bool{ 
                std::cout << xpos << " " << vec.x << " " << vec.y << std::endl;
                return true;
        };
        isBBoxInsideFOVHorizontal(xpos, vec);
    }
    
    return 0;
}

All 3 versions work. Although, this piece of code (local function) should be kept inside the for loop where it belongs conceptually. The for loop is really long and putting this small local function thousands of lines above should be avoided in this case.

Could someone explain clearly if Code 1 (and 3) create multiple structures or has any kind of additional computational burden? Especially in the case of Code 3, keeping FoV (that never changes) in the capture list creates any problem (such as multiple creation of functors)?

OR Codes 1, 2 and 3 are 100% equivalent in compilation/execution terms (of course bear in mind this is just a MWE for discussion and so that you can compile it, actual structures might differ)?