[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20131030103644.GA25479@openwall.com>
Date: Wed, 30 Oct 2013 14:36:44 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: ZedBoard: bcrypt
Katja,
On Wed, Oct 30, 2013 at 01:17:40PM +0400, Solar Designer wrote:
> Also, the swapping of L and R may be avoided by implementing odd and
> even rounds separately. We do this on CPU and it is obviously
> beneficial there, but it may or may not result in overall savings on
> FPGA. You may try both approaches.
Better yet, you may choose L or R on input to each round with a mux,
based on the least significant bit of the round number. In fact, you
only need to apply this towards the half-block (that is, L or R) to be
xor'ed after the S-box lookups. You do not need it for the S-box
inputs, which are always based on whatever was the output of the
previous round (regardless of whether it's L or R).
Alexander
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
