int main(void){ int a[100]; #ifdef _OPENMP #pragma omp parallel for #endif for (int i=0;i<100;i++) { a=i*i; } system("pwd"); return 0; }
This code works if it is compiled without -openmp. But when it is compiled with -openmp, the program stops at the end of the code. This happened when I upgraded from Ubuntu 10.10 to 11.04. This problem didn't occur on Ubuntu 10.10. Also, if I use g++ with -fopenmp, the program works. The compiler version is 12.0.4. I don't know if this is due to the compiler or OS. Does someone have the same problem? Is there a way to solve this issue?
Therefor you do not require the conditional#ifdef _OPENMP around #pragma omp parallel for
If after removing the #ifdef _OPENMP and #endif (leaving #pragma omp parallel for) and the program does not run properly then check to make sure you are linking in the appropriate libraries.
Yes you're right the openmp is double checked once explicitlely bi the ifdef and once implicitely by the pragma. But you're mssing the point. Booby reported a Intel Compiler bug (flag -openmp). You're talking about gcc (flag -fopenmp). He's talking about the system() call, which let's the application hang. You're talking about the "appropriate libraries."
I noticed the same bug yesterday - which led me to this thread, so I can confirm it. It occurs on Ubuntu 11.04 and icc 11.1.069 and 12.0. We never encountered this bug before - even though this is a common use-case here, so I'd say its a Ubuntu 11.04 specific problem.
