3 4 5 6 7 Given is the following C-function that determines the greatest element of a vector arr with size N: i int comp_max(int* arr, int N) 2{ int gmax = arr[0]; for (int i = 1; i gmax) gmax = arr[i]; return gmax; 8} Implement the given function for the use of NVIDIA CUDA such that the greatest element gmax of a vector arr with N Integer elements is determined. Given is the number num blocks of CUDA blocks and the number block_size of threads per block. Assume that N > block_size num_blocks. Conduct the following steps for your implementation: Implement the host function: int comp_max(int* arr, int n, int block_size, int num_blocks) This function should contain the following: - All necessary CUDA memory operations. - The call to the device function with num_blocks and block_size. - The final computation of the global maximum gmax. - The maximum gmax is returned. Assume that all passed parameters are initialized. Implement the device function: void comp_max_device(int* arr, int n, int* grid values) Implement the following phases into the device function: 1. Each thread determines the local maximum tmax of its assigned vector elements. 2. The threads of a block blockIdx.x determine one local maximum for this block in parallel. This value is stored at the position grid.values[blockIdx.x]