One more... Any benefit would be CPU specific: if ((i & 0x1) == 0) rp12 ^= tmppar; Seems like it's ripe for something like: rp12 ^= tmppar & ~(0 - (i & 0x1))

One more... Any benefit would be CPU specific: if ((i & 0x1) == 0) rp12 ^= tmppar; Seems like it's ripe for something like: rp12 ^= tmppar & ~(0 - (i & 0x1)); If my logic is functioning ;-) this eliminates the branch, but many modern CPUs running linux probably have a conditional mov which should be faster with the straightforward version. Edit: Except RISC-V, there are no conditionals other than branch. Maybe this will help on that arch.