1) (10 points) Write a serial implementation to calculate the standard-deviation of an array of randomly generated floating point values. Then thread it using std::threads as well as OpenMP (yes, two implementations!). Compare the performance results (over both N and P). Estimate the overhead costs for both parallelization methods. 2) (10 points) Add a capability to the OpenMP implementation to return a list (values and indices) of all values over a sigma value specified by the user. 3) (5 points) Add the same capability to the std::threads implementation.