Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [day] [month] [year] [list]
Message-ID: <4E3B364E.4020001@linuxasylum.net>
Date: Fri, 05 Aug 2011 02:16:14 +0200
From: Samuele Giovanni Tonon <samu@...uxasylum.net>
To: john-users@...ts.openwall.com
Subject: opencl patch for 1.7.8-jumbo5

hello,
you can find opencl patch for latest john (+jumbo) on the wiki:

http://openwall.info/wiki/_media/john/john-1.7.8-jumbo-5-opencl-1.diff.bz2

This patch fix problems with john-1.7.8 as stated by groszek;
now you have a good excuse to burn your card.

I've added - thanks to the work of Alain - find_best_workgroup
function that will try to detect dinamically best workgroup size
for your card.

there are still some tweaks to do, and i'm planning on trying them
to see if i can make faster crunching even in cards without dma transfer.

Opencl still is not good for "single" mode, due to the large amount
of passwords it need to crunch to stay at a decent speed.

some benchmarks for you from a 5770 hd radeon:

Benchmarking: Raw MD5 [raw-md5-opencl]...
Kernel path is : ./md5_opencl_kernel.cl
OpenCL Platform: <<<ATI Stream>>> and device: <<<Juniper>>>
Max Group Work Size 256 Optimal Group work Size = 128
DONE
Raw:    56062K c/s real, 56062K c/s virtual


Benchmarking: Raw SHA-1 OpenCL [raw-sha1-opencl]...
Kernel path is : ./sha1_opencl_kernel.cl
OpenCL Platform: <<<ATI Stream>>> and device: <<<Juniper>>>
Max Group Work Size 256 Optimal Group work Size = 64
DONE
Many salts:     12713K c/s real, 12588K c/s virtual
Only one salt:  12588K c/s real, 12588K c/s virtual


Benchmarking: Netscape LDAP SSHA OPENCL [salted SHA-1]...
Kernel path is : ./ssha_opencl_kernel.cl
OpenCL Platform: <<<ATI Stream>>> and device: <<<Juniper>>>
Max Group Work Size 256 Optimal Group work Size = 64
DONE
Many salts:     23519K c/s real, 24197K c/s virtual
Only one salt:  17806K c/s real, 19660K c/s virtual


Benchmarking: NT MD4 [OpenCL 1.0]...
Kernel path is : ./nt_opencl_kernel.cl
OpenCL Platform: <<<ATI Stream>>> and device: <<<Juniper>>>
Optimal Group work Size = 1
DONE
Raw:    3563K c/s real, 4475K c/s virtual


this last one: having work size=1 makes me wonder there's still
space to some optimizations.


Regards
Samuele

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.