According to the MinGW64 developers, the math library implementation is faster for gcc. The simulations use openmp which is disabled in MinGW64 due to their buggy pthread implementation which makes openmp hang when used during pthread execution.