Efficient permutation instructions for fast software cryptography
R.B. Lee, Zhijie Jerry Shi, Xiiao Yang · IEEE Micro · 2001
Performing permutations in software can facilitate more widespread use of secure information processing and faster multimedia processing, but current instruction set architectures, even when augmented with subword-parallel multimedia instructions, do not provide efficient, bit-level software permutations. Four new instructions each offer a solution. They are: PPERM (a new, lower-cost version of PPERM3R that selects bits for one byte of the result); GRP (a permutation instruction that separates bits into left and right parts); CROSS (a permutation instruction using Benes interconnection network theory); and OMFLIP (a permutation instruction using enhanced Omega-Flip interconnection network theory).