Parallel FFTs without Inter-Processor Permutations