Further Improvement and a Generalization of Parallel FFTs