Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Wed, 01 Feb 2012 15:43:09 +0100
From: Samuele Giovanni Tonon <samu@...uxasylum.net>
To: john-dev@...ts.openwall.com
Subject: strange behaviour of c/s on raw-md5 and raw-md5-opencl

hello,
i report a strange behaviour on raw-md5 CPU/GPU version.
i've been working on making raw-md5 a bit faster and i succeded
a bit.

../run/john -test -fo=raw-md5-opencl
OpenCL Platforms: 1
OpenCL Platform: <<<NVIDIA CUDA>>> 1 device(s), using device: <<<GeForce
GT 525M>>>
Max Group Work Size 1024 Optimal Group work Size = 512
LWS 512, KpC 2097152
Benchmarking: Raw MD5 [raw-md5-opencl]... DONE
Raw:    49344K c/s real, 24916K c/s virtual

../run/john -test -fo=raw-md5
Benchmarking: Raw MD5 [SSE2i 10x4x3]... DONE
Raw:    24302K c/s real, 24302K c/s virtual

however i'm noticing a strange behaviour on the C/S report:

../run/john -i:digits -fo=raw-md5-opencl /tmp/pppp
OpenCL Platforms: 1
OpenCL Platform: <<<NVIDIA CUDA>>> 1 device(s), using device: <<<GeForce
GT 525M>>>
Max Group Work Size 1024 Optimal Group work Size = 512
LWS 512, KpC 2097152
Loaded 10000 password hashes with no different salts (Raw MD5
[raw-md5-opencl])
guesses: 0  time: 0:00:00:05 DONE (Wed Feb  1 15:07:55 2012)  c/s:
40115M  trying: 82517659 - 83536784

../run/john -i:digits -fo=raw-md5 /tmp/pppp
Loaded 10000 password hashes with no different salts (Raw MD5 [SSE2i
10x4x3])
guesses: 0  time: 0:00:00:06 DONE (Wed Feb  1 15:08:07 2012)  c/s:
185185M  trying: 83536056 - 83536784

this can be seen also using an incremental alpha7 defined as follow :
[Incremental:Alpha7]
File = $JOHN/alpha.chr
MinLen = 1
MaxLen = 7
CharCount = 26



../run/john -i:alpha7 -fo=raw-md5 /tmp/pppp
Loaded 10000 password hashes with no different salts (Raw MD5 [SSE2i
10x4x3])
guesses: 0  time: 0:00:07:13 DONE (Wed Feb  1 15:15:58 2012)  c/s:
192911M  trying: qqjxwsq - qqjxjxq


../run/john -i:alpha7 -fo=raw-md5-opencl /tmp/pppp
OpenCL Platforms: 1
OpenCL Platform: <<<NVIDIA CUDA>>> 1 device(s), using device: <<<GeForce
GT 525M>>>
Max Group Work Size 1024 Optimal Group work Size = 512
LWS 512, KpC 2097152
Loaded 10000 password hashes with no different salts (Raw MD5
[raw-md5-opencl])
guesses: 0  time: 0:00:05:12 DONE (Wed Feb  1 15:21:51 2012)  c/s:
48408M  trying: qqhjueq - qqjxjxq

same password file, same incremental mode on CPU it last for 7:13 at
at 192911M c/s , on GPU it last for 5:12 at 48408 M c/s ? should'nt
the c/s be higher ?

this has been tested on last github and on john-1.7.9-jumbo5
i'm including source code so anyone can test and confirm it or not.

Cheers
Samuele


View attachment "rawMD5_opencl_fmt.c" of type "text/x-csrc" (14132 bytes)

Powered by blists - more mailing lists

Your e-mail address:

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