C: pointer to a function that returns pointer to an integer

Question

I'm trying to write a code that evaluates the run-time of the search algorithm, and I have doubts how to do it correctly. I want to pass a function linear_search to the function code_timer, and code_timer function is expected to evaluate the execution time of the linear search function. In order to solve this problem, I'm trying to use a pointer to a function.

// Search Algorithm function prototype
int *linear_search(int arr[], int arr_size, int val);

// Code timing function that evaluates the performance of the algorithm function
double code_timer( int *arr, int arr_size, int val, int nrun,
        int* (*search_algorithm)(int *arr, int arr_size, int val) )
{
    double interval = 0;
    clock_t start, end;
    int *ptr = NULL;

    start = clock();
    /* Code under Test */
    for (int j = 0; j < nrun; j++) {
        ptr = (*search_algorithm)(arr, arr_size, val);
    }
    /* End of Code under Test */
    end = clock();
    interval = (double)(end - start) / (double)(nrun) / (double) CLOCKS_PER_SEC * 1000000.0;
    return interval; // in microseconds (mcs)
}

Finally I need to call the function code_timer with the function linear_search as an argument.

#define SIZE 100000 //size of data array
#define NRUN 100 // number of runs

int main(void) {
    ...
    // calling code_timer with a function linear_search as a parameter
    linear_result = code_timer(benchmark_data, SIZE, check_point, NRUN, linear_search);
}

I would like to know if I'm doing it correctly, and if no, please advise the way I could do it. Thank you!

Update: My main concern is pointer to function:

1. Am I declaring it correctly as an argument?

double code_timer( int *arr, int arr_size, int val, int nrun,
        int* (*search_algorithm)(int *arr, int arr_size, int val) );

2. Am I dereferencing it correctly in code_timer and main functions?

// this is from code_timer function
    for (int j = 0; j < nrun; j++) {
        ptr = (*search_algorithm)(arr, arr_size, val);
    }

// this is from main function
    linear_result = code_timer(benchmark_data, SIZE, check_point, NRUN, linear_search);

Answer 1

No this isn't done correctly. One obvious problem is that you can't have printf or any other I/O in the middle of the benchmarking code. It's gonna be the main bottleneck, grabbing all the execution time each lap of the loop. So this code is just benchmarking printf performance + some sugar on top.

Also, printf("Found value: %d\n", *ptr); outside the for loop is strange, surely a bug?

This is what you should do instead:

• Start the clock
• Run the function n times in a loop.
• Store the results in a pre-allocated array.
• Stop the clock.
• Print algorithm results.
• Print benchmarking results.

Preferably use a static array, since dynamic memory access can lead to late OS allocation, meaning the OS might execute the allocation during the first time you actually access the data, rather than at the line where the malloc call is located. And that would screw up benchmarking completely.