Implementation complexity of bit permutation instructions

Several bit permutation instructions, including GRP, OMFLIP, CROSS, and BFLY, have been proposed recently for efficiently performing arbitrary bit permutations. Previous work has shown that these instructions can accelerate a variety of applications such as block ciphers and sorting algorithms. In this paper, we compare the implementation complexity of these instructions in terms of delay. We use logical effort, a process technology independent method, to estimate the delay of the bit permutation functional units. Our results show that for 64-bit operations, the BFLY instruction is the fastest among these bit permutation instructions; the OMFLIP instruction is next; and the GRP instruction is the slowest.