Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Thu, 13 Sep 2012 22:37:28 +0200
From: magnum <john.magnum@...hmail.com>
To: "john-dev@...ts.openwall.com" <john-dev@...ts.openwall.com>
Subject: office-opencl

I have committed a first version of office-opencl format. For now, it only supports Office 2007 documents, and a max plaintext length of 19. I could re-use my SHA-1 from RAR, and the iterations are MUCH more GPU-friendly here:

Using device 0: GeForce GTX 570
Benchmarking: Office 2007 SHA-1 AES [OpenCL]... DONE
Raw:    12136 c/s real, 12136 c/s virtual

Using device 0: Tahiti
Benchmarking: Office 2007 SHA-1 AES [OpenCL]... DONE
Raw:    40832 c/s real, 748982 c/s virtual

Each key is 50004 x SHA-1. The nvidia does about 605 million sha1/s while the AMD exceeds 2 billion. This can be tweaked a little more, but I'll do 2010/2013 first.


The format even works on a MacBook with Apple's über-buggy driver:

Using device 1: GeForce GT 650M
Benchmarking: Office 2007 SHA-1 AES [OpenCL]... DONE
Raw:	2073 c/s real, 273066 c/s virtual


As a reference, this is the CPU format using the same one 2007 test vector and running on one core on Bull:

Benchmarking: Office 2007/2010 (SHA-1) / 2013 (SHA-512), with AES [32/64]... DONE
Raw:    115 c/s real, 115 c/s virtual

BTW the CPU format could very easily be adopted to use SSE2 for near 4x speedup per core, although the flow needs to be altered a little just like in the OpenCL format. I will concentrate on OpenCL though.

magnum

Powered by blists - more mailing lists

Your e-mail address:

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