[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20120613180221.GA24031@openwall.com>
Date: Wed, 13 Jun 2012 22:02:21 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: bf-opencl
Sayantan -
I've just tried running bf-opencl in magnum-jumbo on HD 7970 against
pw-fake-unix from:
http://openwall.info/wiki/john/sample-hashes
This triggered an ASIC hang on the 7970 after a few minutes. Also, only
one password was cracked:
$ ./john -fo=bf-opencl ~/john/pw-fake-unix -w=password.lst -pla=1
OpenCL platform 1: AMD Accelerated Parallel Processing, 2 device(s).
Using device 0: Tahiti
*****See 'opencl_bf_std.h' for device specific optimizations******
Loaded 3107 password hashes with 3107 different salts (OpenBSD Blowfish OpenCL [BF_OPENCL])
Suppressed 1 duplicate lines.
omega (u2461-bf)
guesses: 1 time: 0:00:00:14 0.00% c/s: 233 trying: 123456 - jethrotull
guesses: 1 time: 0:00:00:21 0.00% c/s: 457 trying: 123456 - jethrotull
guesses: 1 time: 0:00:00:25 0.00% c/s: 519 trying: 123456 - jethrotull
guesses: 1 time: 0:00:00:29 0.00% c/s: 565 trying: 123456 - jethrotull
guesses: 1 time: 0:00:00:33 0.00% c/s: 601 trying: 123456 - jethrotull
guesses: 1 time: 0:00:00:37 0.00% c/s: 629 trying: 123456 - jethrotull
guesses: 1 time: 0:00:00:40 0.00% c/s: 652 trying: 123456 - jethrotull
guesses: 1 time: 0:00:00:44 0.00% c/s: 671 trying: 123456 - jethrotull
guesses: 1 time: 0:00:00:48 0.00% c/s: 687 trying: 123456 - jethrotull
guesses: 1 time: 0:00:00:52 0.00% c/s: 701 trying: 123456 - jethrotull
guesses: 1 time: 0:00:00:55 0.00% c/s: 713 trying: 123456 - jethrotull
guesses: 1 time: 0:00:01:33 0.00% c/s: 779 trying: 123456 - jethrotull
guesses: 1 time: 0:00:01:49 0.00% c/s: 793 trying: 123456 - jethrotull
Somehow the c/s rate starts low and increases slowly. I don't have a
good explanation for this in this case. The benchmark had reported
something like 2500 c/s.
Running on GTX 570 cracked only this same omega password (I interrupted
after a minute or so).
I realize that matching 3000 plaintexts against 3000 hashes (each with
its own salt) at this low c/s rate is very time-consuming, and is a bit
more time-consuming when keys_per_crypt is large, but I think we should
be seeing many more passwords cracked during 2 minutes. Each
crypt_all() call should result in about 1 password cracked, and it
should be taking about 4 seconds (at 800 c/s). So we should have seen
about 25 passwords cracked above, not one. I think there's a bug.
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.
