What is the exact meaning of objective function evaluations in case of a curve fitting?

Hello, I need to figure out the meaning of a function evaluation when I do a curve fitting. Trust-region-reflective and Levenberg-Marquardt algorithms are returning the total number of objective function evaluations and this quantity is a measure of robustness of the used method. Let's say my function is: f(x,y) = a*exp(x)-y. I want to find "a" by curve fitting. I have the following data set: x=[1,2,3], y = [2.1,44,0.22] What is the meaning of one function evaluation in this case? The evaluation of the function in one data point (let's say [1,2.1]) or the evaluation of the function in all three data points and obtaining the vector of three elements? I am trying to find out how many instructions my curve fitting needs, when I realize the curve fitting in a processor. Thank you for your help in advance.