Why every time getting different output in openmp code?

Question

Every time I execute cal() function with the same parameter I get different output. Function g() always calculate the same result for same input. Are threads overwriting any variable?

void cal(uint_fast64_t n) {
    Bint num = N(n);
    Bint total = 0, i, max_size(__UINT64_MAX__);
    for(i = 1; i <= num; i+= max_size){
        #pragma omp parallel shared(i,num,total)  
        { 
            int id = omp_get_thread_num();
            int numthreads = omp_get_num_threads();
            Bint sum(0), k;
            for(uint64_t j = id; (j < __UINT64_MAX__); j+=numthreads){
                k = i+j;
                if(k > num){
                    i = k;
                    break;
                }
                sum = sum + g(k);
            }
            #pragma omp critical
                total += sum;
        }
    }
    std::cout << total << std::endl;
}

`for(uint64_t j = id; (j < __UINT64_MAX__); j+=numthreads){` probably does not do what you want. The loop conditions is always true unless `j == __UINT64_MAX__`, which is impossible for e.g. even `id` and `numthreads`. But you have UB regardless so this is only secondary. — Max Langhof, Sep 03 '19 at 15:02
Its BigInt object to hold large numbers (>2^64). I am using [this](https://github.com/Maritimo/Bint.git) library. — Ajit Jadhav, Sep 03 '19 at 15:12
Actually I am trying to solve this [this](https://projecteuler.net/problem=672) problem programmatically. I computed H(10) but to solve H(10^9) uint64_t is also not sufficient.That's why I am using BigInt. — Ajit Jadhav, Sep 03 '19 at 15:19

Max Langhof · Answer 1 · 2019-09-03T14:53:02.677

7

if(k > num){
    i = k;
    break;
}

Here you modify the shared variable i (possibly multiple times in parallel) while other threads may be reading from it (for k = i+j), all without synchronization. This is a race condition and your code thus has Undefined Behavior.

edited Sep 03 '19 at 14:53

answered Sep 03 '19 at 14:47

Max Langhof

23,383
5
39
72

This is my single thread function: ```void cal_single_thread(uint_fast64_t n) { Bint num = N(n); Bint sum = 0, i = 0; for(i = 1; i <= num; ++i){ sum += g(i); } std::cout << sum << std::endl; }``` How can i modify it for openmp ? – Ajit Jadhav Sep 03 '19 at 15:58
@AjitJadhav There are several different ways, but comments are not the appropriate place to show those. In any case, I **very** much doubt that you will ever reach `i >= __UINT64_MAX__`. Do you know how large 2^64 is? 2^64 nanoseconds (~= CPU cycles) take **600 years**. – Max Langhof Sep 03 '19 at 16:02
Haha :).. you are right, I have to think in a different way to solve that! Anyway, due to that problem, I learned many things. Thank you all. I was surprised by the quick replies. – Ajit Jadhav Sep 03 '19 at 16:37

score 3 · Answer 2 · answered Sep 03 '19 at 14:46

The value of j depends on the value of id. If different threads are used to do the math, you'll get different results.

        int id = omp_get_thread_num();      // <---
        int numthreads = omp_get_num_threads();
        Bint sum(0), k;
        for(uint64_t j = id; (j < __UINT64_MAX__); j+=numthreads){  // <---
            k = i+j;
            if(k > num){
                i = k;      // <---
                break;
            }
            sum = sum + g(k);

Further, you change i to k when k > num. This can happen much sooner or much later depending on which thread is picked up first to run the inner loop.

You may want to look at this question and answer.

Does an OpenMP ordered for always assign parts of the loop to threads in order, too?

Why every time getting different output in openmp code?

2 Answers2