Why openmp 32 thread is much slower than 1 thread?

Question

I am trying to write an application calculating l2 norm of 2 arrays. I have to parallel my calculation.

Here is my paralelled code:

   double time_star