john-users - Re: john.pot problem

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sat, 14 Jan 2012 12:17:21 -0500
From: David LaPorte <david@...idlaporte.org>
To: john-users@...ts.openwall.com
CC: Solar Designer <solar@...nwall.com>
Subject: Re: john.pot problem

> BTW, the c/s rate is quite low for OpenCL.  What hardware, OS, OpenCL
> implementation is that?
> 
> For example, you can obtain a c/s rate like this (for the same hash type
> indeed) on a Core 2 Duo by building -jumbo with the linux-x86-64i make
> target and enabling OpenMP in the Makefile - that's without OpenCL.

It did seem low to me as well.  Here is the status after running
1.7.9-jumbo5 with MPI for 1hr executed with:

mpiexec -np 6 ./john --incremental /home/dlaporte/hashes

guesses: 0  time: 0:01:00:00 0.00%  c/s: 17570
guesses: 0  time: 0:01:00:00 0.00%  c/s: 17605
guesses: 0  time: 0:01:00:00 0.00%  c/s: 17760
guesses: 0  time: 0:01:00:00 0.00%  c/s: 17726
guesses: 0  time: 0:01:00:00 0.00%  c/s: 17436
guesses: 0  time: 0:01:00:00 0.00%  c/s: 17420

Here is the status after 1hr with 1.7.9-allopencl-04 executed with:

./john --incremental /home/dlaporte/hashes

guesses: 0  time: 0:01:01:10  c/s: 66062

It's an HD5830 with a Phenom II X6 1055 and 16GB RAM running Fedora 16
(x86_64).  I'm using AMD-APP-SDK-v2.6-RC3-lnx64 and the 11.12 driver.
Does these numbers seem in-line with what I should expect?

Thanks for the "unique" tip, I'll try optimizing everything once I get
this figured out.

Dave