Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [day] [month] [year] [list]
Date: Sun, 19 Jul 2015 23:07:23 -0400
From: Alain Espinosa <alainesp@...ta.cu>
To: john-dev@...ts.openwall.com
Subject: RE: faster way to transpose matrix of words for sse and
 bits for bitslice



-------- Original message --------
From: Aleksey Cherepanov <lyosha@...nwall.com> 
Date:07/19/2015 4:50 PM (GMT-05:00) 
To: john-dev@...ts.openwall.com 
Cc: 
Subject: [john-dev] faster way to transpose matrix of words for sse and bits for bitslice 

...It seems that transposition is a common operation in john: sse needs
to transpose matrix of integers and bitslice needs to transpose a
matrix of bits.

...I found an algo to transpose bit matrices faster than bit by bit in
Hacker's Delight.

You can take a look at my bitslice SHA256 code. There I use Hacker's Delight algo modified to use SIMD intrinsics to transpose 32 bit matrixs.

...I think current descrypt and LM uses something similar. LM could
transpose the matrix back and remove FMT_BS flag because we expect
many hashes (right?) unlike with salted descrypt.

LM may give a good speedup, descrypt not so much. Some time ago I check John LM setkey code and if I understand it correctly it doesn't use bit transposing, so we can improve it.

Regards,
Alain
[ CONTENT OF TYPE text/html SKIPPED ]

Powered by blists - more mailing lists

Your e-mail address:

Powered by Openwall GNU/*/Linux - Powered by OpenVZ