2) Implement two versions of a summation algorithm discussed in Parallel Algorithm Design using OpenMP .The first version must implement the PRAM binary-tree summation using only: #pragma omp parallel o omp get_num_threads0 o omp_get thread num0 o omp_set_thread_num0 you CANNOT use the for reduction call. The second version should implement a partitioning based summation algorithm. For this implementation, you can use everything taught so far. Use this as the summation routine declaration: @brief Calculate the sum of an array @param pd_input -input pointer to array of data to be summed @param i_N -input size of array to be summed *@return sum of array double fast_sumidouble "pd_input, int i_N); Assume the size of the array and the number of processors available are independent variables (ieTV can be really big and p' can be less than optimal). .Benchmark the routine for various sizes of N and various values of p. Relate the results to Amdahl's Law and Brent's Theorem 2) Implement two versions of a summation algorithm discussed in Parallel Algorithm Design using OpenMP .The first version must implement the PRAM binary-tree summation using only: #pragma omp parallel o omp get_num_threads0 o omp_get thread num0 o omp_set_thread_num0 you CANNOT use the for reduction call. The second version should implement a partitioning based summation algorithm. For this implementation, you can use everything taught so far. Use this as the summation routine declaration: @brief Calculate the sum of an array @param pd_input -input pointer to array of data to be summed @param i_N -input size of array to be summed *@return sum of array double fast_sumidouble "pd_input, int i_N); Assume the size of the array and the number of processors available are independent variables (ieTV can be really big and p' can be less than optimal). .Benchmark the routine for various sizes of N and various values of p. Relate the results to Amdahl's Law and Brent's Theorem