Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 9 Jul 2012 05:39:58 +0200
From: Lukas Odzioba <lukas.odzioba@...il.com>
To: john-dev@...ts.openwall.com
Subject: Re: raw-sha256-cuda (was: [PATCH] integrate psk-crack
 (from ike-scan) into john.)

2012/7/8 Solar Designer <solar@...nwall.com>:
> On Sat, Jul 07, 2012 at 07:19:30PM +0200, Lukas Odzioba wrote:
>> Just ~8M:
>> ukasz@...l:~/current-magnum/ukasz-jumbo/src$ ../run/john -test
>> -fo=raw-sha256-cuda
>> Benchmarking: raw-sha256-cuda [SHA256]... DONE
>> Raw:    7831K c/s real, 7754K c/s virtual
>>
>> 6.5M on gtx460
>>
>> But on wiki there is info about 10k version of this patch, and I am
>> absolutely sure that I was working on it this year.
>> I must dig through my hdd to find a proper code.
>>
>> Anyway 10M or even 15M is still slow.
>
> Yes, it's weird that your raw SHA-256 code is so slow.  It doesn't even
> hit the formats interface and CPU to GPU transfers bottleneck.  IIRC,
> myrice's SHA-512 code performed faster even with a similarly large
> PLAINTEXT_LENGTH.
>
> BTW, "raw-sha256-cuda [SHA256]" suggests that you're using a code
> revision different from what's in magnum-jumbo, which has:
>
> #ifdef SHA256
>   #define FORMAT_LABEL          "raw-sha256-cuda"
>   #define FORMAT_NAME           "Raw SHA-256"
> ...
> #endif
> #define ALGORITHM_NAME          "CUDA, unreliable, may miss guesses"
>
> And we need to actually make it reliable, so that we can remove the
> "unreliable" notice.
>
> Alexander

After some digging I found a bit better code 16.5M (gtx 460) from april.
I'll test it and if it works we might want it in repo.

Lukas

Powered by blists - more mailing lists

Your e-mail address:

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