[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <4EEF7B47.7090901@hushmail.com>
Date: Mon, 19 Dec 2011 18:58:31 +0100
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: 1.7.9-jumbo
On 12/19/2011 01:02 PM, Solar Designer wrote:
> On Sun, Dec 18, 2011 at 10:56:34PM +0100, magnum wrote:
>> How can a dropped conditional make a significant performance drop?
>
> My guess is that the condition actually affected the number of loop
> iterations in some cases that are triggered in some(?) benchmark runs.
> If this differs between builds, then I'd suspect that an uninitialized
> value is involved.
There are no uninitialized variables in the NT code if that is what you
meant. Anyway, it turns out the version that does truncate correctly
regained its speed just by changing the order of conditions, ie. from
if (md4_size < (PLAINTEXT_LENGTH - 1) && (temp = key[++md4_size]))
to
if ((temp = key[++md4_size]) && md4_size < PLAINTEXT_LENGTH)
Please try the version at
https://github.com/magnumripper/magnum-jumbo/raw/bfb40805651e53ea4cd8b02ef00fcf4056763061/src/NT_fmt_plug.c
and see how this performs on your gear compared to 1.7.8-j8.
It's actually 2% faster on my laptop so I'd be surprised if this is not
the definite fix.
magnum
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.
