Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [day] [month] [year] [list]
Date: Sun, 15 Jan 2012 02:39:23 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: loader speed

Hi,

Here are some JtR loader speed numbers (on the E5420 box I am using for
JtR speed tests lately):

1.7.9.2 from CVS:

1 million LM hashes - 1.85 sec
10 million LM hashes - 27.2 sec

1.7.9-jumbo-5 with PASSWORD_HASH_THRESHOLD_* taken from 1.7.9.2 above:

1 million LM hashes - 2.1 sec
10 million LM hashes - 29.6 sec

Both with --format=lm forced explicitly, so no extra valid() calls (for
formats only present in jumbo) should have been involved.  Both with
empty john.pot.

I was measuring with commands like:

echo | time ./john --format=lm --stdin passwd

So jumbo's loader appears to be about 10% slower.  I did not investigate
why.  Maybe because of the introduction of prepare()?

BTW, the current choice of PASSWORD_HASH_SIZE_FOR_LDR = 4 seems optimal.
While going for 5 reduces the load time for 10 million hashes by about
50% (to 18 sec), it also increases load time for 1 million hashes by 70%
(to over 3 sec).  And it's extra memory that may or may not be returned
to the OS.  So unless we make it conditional on expected hash count (by
file size?), I think we should just keep it at 4.

Alexander

Powered by blists - more mailing lists

Your e-mail address:

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