Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <46979A5A.2020101@banquise.net>
Date: Fri, 13 Jul 2007 17:29:30 +0200
From: Simon Marechal <simon@...quise.net>
To: john-users@...ts.openwall.com
Subject: Re: QS20 Cluster

James Evans wrote:
> I now have access to a 14 node Cell cluster (IBM QS20s). If anyone would like to get to work on porting JtR to get some nice performance improvements with the SPEs, let me know!
> 
> James

Based on my tests, you will have the utmost difficulties porting john to
the SPUs, and it is VERY likely it will prove impossible (just look at
the 256k memory limitation for both code & data ...).

Another approach is to just write the hashing functions for the SPUs.
The performance gain will be nullified by the speed of DMA / mailbox
accessess.

A hybrid approach would be to write part of the password generator in
the SPUs and run the hashing function on them. This would prove fast,
but very far from the john source. (If you have a good idea on how to
solve the compare problem with many passwords, I'll be glad to learn it!)

I do believe the proper way to crack passwords on this processor is to
write a dedicated program.

-- 
To unsubscribe, e-mail john-users-unsubscribe@...ts.openwall.com and reply
to the automated confirmation request that will be sent to you.

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.