I have a question about the precision of floating point operations while using openmp. I solve a large sparse matrix.
Doing this without openmp, the code works. Running it several times the results are identical each time. But when I compile it with openmp, the results are not identical. This is even the case, when I remove all !$OMP directives, so I shurely have no race conditions.
And the results differ between a computation without using openmp and a computation with using openmp and using one thread only.
I use the following oponmp-options:
I guess, the numerical precision is set different when using openmp and rounding errors occour or something else. In the documentation are several compiler options for setting the numerical behavior, but this didnt help me. Setting /fp:precise has no effect.
Do you have an idea or some advice for me?
Thank you and best regards,