I am trying to multiple the values of a matrix. I have both a parallel and sync version. But the parallel version is longer than the sync version. Current the sync takes around 90 seconds and the parallel over 100. Which makes no sense to me. My logic was to split the matrix into 4 parts from the first 4
Tag: openmp
How to control global OpenMP settings from C/C++?
OpenMP doesn’t provide API to adjust some crucial settings in runtime, e.g. OMP_WAIT_POLICY or GOMP_SPINCOUNT (at least in gcc 4.8.3). Such settings are designed to be changed through environment variables for some reason. Not very convenient. Is there any way I can change such settings in run-time from C/C++ using gcc on Linux? For example, would it be enough to