The "gzip flip" stages perform a permutation that effectively reverses the order of the inner stages. This enables one network to perform the shuffle and unshuffle operations. Otherwise we'd need two networks, one for shuffle and one for unshuffle.
-
-
Show this thread
-
You can easily find work regarding shuffle operations *used* in permutation networks from 50+ years ago. But I can not find any research regarding permutation networks that can implement a generalized form of shuffle/unshuffle operations. Pointers are very welcomed!
#asktwitterShow this thread -
Btw, those "gzip" operations are like Inception for bit permutations: You permute the bits in your data word by permuting the bits in the bit indices. This network (used repeatedly) can performs a class of 32-bit permutations that is equivalent to all 5-bit permutations.
Show this thread
End of conversation
New conversation -
-
-
This Tweet is unavailable.
-
Please ignore the chapter on the "shuffle" instruction. It will be gone by the end of the week. :)
End of conversation
-
-
-
Very useful for some cryptographic algorithms...
-
Well.... DES. Most (all?) newer crypto algorithms avoid bit permutations because CPUs are so bad at them. :) Maybe XBitmanip ends up becoming a RISC-V standard, and RISC-V ends up becoming the mainstream platform, then we might see crypto using permutations more again. :D
- Show replies
New conversation -
-
-
The flip stages look like a bit-reverse. Each intermediate stage looks like a simple bit-index-permutation. For example, take a 5 bit index of the form [a,b,c,d,e] and map it to [a,c,b,d,e]. So element 11001 is mapped to element 10101.
-
Yes! :D I just tweeted this (before I saw your tweet).https://twitter.com/oe1cxw/status/988494996834848768 …
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.