Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 16 Jul 2012 10:55:55 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: xsha512-cuda & xsha512-opencl testing

myrice -

As discussed on IRC, you'll need to redo these with incremental mode and
with fewer passwords getting cracked.  However, thank you for posting
these "wrong" results as well.  It helps.  They're also good for
correctness testing before you approach actually benchmarking things.

On Mon, Jul 16, 2012 at 02:29:40PM +0800, myrice wrote:
> HashNum_SaltNum
> 1_1
> guesses: 1  time: 0:00:00:01 DONE (Sun Jul 15 12:09:50 2012)  c/s: 499321
> 
> 100_100
> guesses: 100  time: 0:00:00:01 DONE (Sun Jul 15 18:18:18 2012)  c/s: 26346K

1 second is too little time for the c/s rate to reflect anything.
(1 minute could be OK.)

> 10K_10K
> guesses: 10000  time: 0:00:01:36 DONE (Sun Jul 15 19:25:42 2012)  c/s: 54613K

This looks reasonable and consistent with --test, 54.6M c/s at "many salts".

> 10K_100
>  guesses: 10000  time: 0:00:00:03 DONE (Sun Jul 15 19:27:50 2012) c/s: 2569M

This could also be reasonable, but the hashes/salt ratio was changing
(so we can't reliably deduce the hashes computed per second from the
printed effective c/s rate) and the total running time is too short.

We need 1+ minute and no successful guesses.

> 10K_1
> guesses: 10000  time: 0:00:00:01 DONE (Sun Jul 15 19:30:25 2012 )  c/s: 5242M

Same problems as above.

> 1M_K
> guesses: 1000000  time: 0:00:10:14 DONE (Sun Jul 15 20:30:56 2012)  c/s: 1260M1M

Running time is OK, but the hashes/salt ratio was changing, and
non-negligible time might have been spent on processing the successful
guesses.

Also, what's "1260M1M" - doesn't look like a number to me.  Is this what
john actually printed?  I doubt it.

> Here is not finished result of 1M hashes and 1M salts
> 1M_1M
> guesses: 73694  time: 0:01:04:35 12.73% (ETA: Mon Jul 16 05:05:37
> 2012)  c/s: 49966K

Now this is actually relevant, similarly to your 10K_10K run.  So things
look OK for the one hash per salt case.

Thanks,

Alexander

Powered by blists - more mailing lists

Your e-mail address:

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