Re: Optimized S-Boxes in Software



Tom St Denis <tom@xxxxxxx> writes:
Even if you were using SSE and doing 128 substitutions in parallel
(block size of 128x8=1024 bits say) you'd still have probably
100-200 SSE operations per substitution. It'd really not come out ahead.

Maybe it will get more interesting with Larrabee, which will have 32
vector registers of 512 bits each, if I understand it right.
.



Relevant Pages

  • Re: Schneiers "Helix" cipher is remarkably similar to the "generic feistel cipher&qu
    ... Tom St. Denis wrote: ... a cipher to the design of simpler components, ... the cipher answered to by Wagner ...
    (sci.crypt)
  • Re: XORShift PRNG as a diffusion structure
    ... >>like him and Tom St Denis. ... > This isn't the first time you've done this before [or the first time ... number of general readers would certainly have been angry ...
    (sci.crypt)
  • Re: AES trickery ;-)
    ... Too bad you missed the day in school where they taught you to ... Someone who reads the above diatribe about how Tom St Denis demands ... you're too american to sit down and work out the details ...
    (sci.crypt)
  • Re: Working on ARC4-16 bit
    ... Well, i'm sorry about what happened with Tom St Denis, i have esteem ... of him and personally i had found interesting many of his posts here. ... deeply tested by highly qualified matematicians and cryptoanalists it ...
    (sci.crypt)
  • Re: LibTomCrypt [v0.96 bug fix]
    ... This is only your imagination, because I dont listen to YOU, Tom St ... Denis, who failed to produce ANY valuable on-merit comment on VMPC. ... You routinely ignore peoples advice, ...
    (sci.crypt)