OpenMP 不产生额外的线程

Question

I'm new to OpenMP and cannot for the life of me utilise multiple threads. I have my environment variable set. Here is a snippet of code which simply should just iterate through the mandelbrot set:

#include <omp.h>
#include <limits>

#define WIDTH 10000
#define HEIGHT 10000
#define INFINITY 2.0f
#define ITERATIONS 1000

using namespace std;


int main()
{

    #pragma omp parallel for
    for (size_t py = 0; py < HEIGHT; py++) {
        for (size_t px = 0; px < WIDTH; px++) {

            float x0 = -2.5f + (px * (1.0f - -2.5f) / WIDTH);
            float y0 = 1.0f + (py * (-1.0f - 1.0f) / HEIGHT);

            unsigned short iteration;
            float x = 0.0f;
            float y = 0.0f;
            for (iteration = 0; iteration < ITERATIONS; iteration++) {
                float xn = x * x - y * y + x0;
                y = 2 * x * y + y0;
                x = xn;
                if (x * x + y * y > INFINITY) {
                    break;
                }
            }
        }
    }
}

Whenever I run this, it never spawns additional threads. I feel I'm doing something horribly wrong. Any help would be appreciated, thanks.

Answer 1

I needed at add the flag -fopenmp to my compiler arguments. Now works properly.

Answer 2

Just to add to the missing -fopenmp flag, OpenMP supports different work scheduling modes and the simple equal work distribution may not reach high efficiency on the run-time minimization for unbalanced workloads like Mandelbrot-Set generation.

#pragma omp parallel for schedule(dynamic)
for(...){}

can be faster for unknown work-intensity per pixel with speedup depending on the unbalance between pixels.