Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20120805003939.GA12412@openwall.com>
Date: Sun, 5 Aug 2012 04:39:39 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: Lukas status report #14

On Sun, Aug 05, 2012 at 02:08:40AM +0200, Lukas Odzioba wrote:
> Last week:
> I spent some time on testing code before contest to avoid any
> "problems", and also helped a bit in password cracking.
> I worked on making md5crypt-opencl code faster. For my 5850 it is 2x
> faster now (150k -> 360k - still slow it should be ~1M)
> For gtx460: ~500k
> For gtx570: ~1M c/s
> For 7970: ~2.2M c/s

It's a pity we didn't have this during the contest.  The bug you mention
below would make the code almost unusable, though, given that there were
mostly long passwords/phrases in the contest.

Why is the code so slow on your 5850?

> Now adding data to buffers is realized like that:
...

Shouldn't you be doing this before the 1000 iterations loop, thus making
it non-performance-critical (and maintaining several buffers - IIRC, you
said you needed 8 of them)?

> There is also a bad news. I have a bug that is pretty hard to locate.
> Code works only for passwords shorter that 8 chars.
> I do not see a reason for that, I must compare step by step execution
> with comparision to other code.
> I'll make proper fixes before end of gsoc.

Shouldn't this be among your plans for this week? -

> This week:
> des-crypt on gpu.
> Any ideas, links, filenames, pdfs to read are welcomed.

I'm afraid the primary reading material would be DES_bs_b.c.

If we had decent support for fast hashes on GPU, you could start by
implementing the simpler LM hash first, but since we don't (and since
it's you rather than myrice, whose project is fast hashes), starting
with descrypt actually makes sense.

On the other hand, if/since your initial goal is to get the code working
rather than to get it working fast, you may start with LM hashes - fully
knowing that you won't outperform the CPU code because of the bottleneck
(solving it would be a task for myrice then).  Once you get this
working, you could proceed with descrypt.

This choice is up to you.

Thanks,

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.