https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65589
--- Comment #4 from felix.ospald at gmx dot de --- (In reply to Jakub Jelinek from comment #3) > Can't reproduce here, with neither 4.8 nor 4.9. But tested just on 16 > threads box (both with argument 16 and 32). With only 16 threads I also cannot reproduce the error (I think it is just more unlikely that somehow two threads infere). Also it sometimes takes several tries to reproduce the error. > Note rand () is not thread-safe, so calling it from multiple threads > concurrently is undefined-behavior. Yes I know, but I can reprodce the error also with the omp critical pragma, but it takes longer. > > Haven't spotted anything wrong in the generated code. Thank you. I will compile the trunk version and try again.