Question

I have alike C++ program with openMP usage

#pragma omp parallel for private(i)
for (j=0;j<NUM_STEPS_J) {

    for (i=0;i<NUM_STEPS_I;i++) {

        std::cout << "Print some information about step i" << std::endl;


    }

    std::cout << "Check of item " << j << " finished" << std::endl;
}

What is the best way to provide a correct output in my case?

I know, that using “printf” instead “cout” solves this problem.

But when I changed “cout” to “printf” the time of my program’s execution increased from about 80 seconds to about 120 seconds.
I think, it is a sufficient influence on program’s productivity.

What is the best way to solve this problem without “printf”?

Is it possible to lock “cout” function during output in some way?

Answer 1

I would be surprised to see printf being slower than iostreams… (and more so considering that you are using std::endl that forces flushing of the buffer) but at any rate, you can use a stringstream to construct the output at once and then call either printf or std::cout << once with the already built line.