[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20120802164243.GB4363@openwall.com>
Date: Thu, 2 Aug 2012 20:42:43 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: Result of hard core password generation on 7970
On Fri, Aug 03, 2012 at 12:32:23AM +0800, myrice wrote:
> Here is bitmaps result, currently I only use global memory.
>
> 1: guesses: 0 time: 0:00:01:22 0.00% c/s: 758186K ~= 758M c/s
Are you using the bitmap even with just 1 hash loaded? In the real
implementation, I guess you'll need to do direct comparisons when the
number of loaded hashes (for the current salt if applicable) is below a
threshold (like just 1 or 2).
> 1000: guesses: 0 time: 0:00:01:16 0.00% c/s: 515938M ~= 515M c/s
> 1M:guesses: 0 time: 0:00:01:51 0.00% c/s: 353254G ~= 353M c/s
OK, this is an improvement upon what you had before.
> For further performance gains. I am trying to use local memory when
> the bitmaps size is less than 8KB. I am debugging this(load bitmaps
> from global memory to local memory seems not work, still
> investigating).
As an option, you could form the bitmap in local memory right away,
based on the list of loaded hashes for the current salt. I guess you'd
have to redo it for every crypt_all() call, though.
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.
