[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Fri, 19 Apr 2013 15:06:50 -0500
From: "jfoug" <jfoug@....net>
To: <john-dev@...ts.openwall.com>
Subject: RE: SHA-2 in sse-intrinsics.c
Very Good!
Trying to squeeze too much out of the raw format, can make it very hard to
actually use the code. Especially when trying to squeeze things through
compiler specific optimizations.
We should also try to get this to use a 'rolling' 16 feedback variables
(like I did with SHA1). I will have to give that a go this weekend.
Reducing the size of this 'overhead' part of the crypt did well in SHA1. I
am pretty sure the same thing can be done on the SHA2, since the feedback
variables are all offset(mod 16) from the current word being worked on.
From: magnum Sent: Friday, April 19, 2013 14:38
>
>I copied Jeremi's SHA-256 and SHA-512 functions from his raw formats, to
sse-intrinsics.c before we start optimizing things away in the formats :-)
>
>It's currently ifdef'ed out and not complete anyway.
>
>We should consider what interface(s) we need for the shared functions.
Maybe we should have a simple interface for non-iterated use (scalar input),
plus one more like our existing md4/5/sha1 (vector input) that is optimized
for iterated use.
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.
