Parallel optimization with boundary elements and kernel independent fast multipole method