john-users - Re: Is JTR MPIrun can be optimized for more cores ?

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 10 Mar 2010 01:09:45 +0100
From: "Magnum, P.I." <rawsmooth@...dband.net>
To: john-users@...ts.openwall.com
Subject: Re: Is JTR MPIrun can be optimized for more cores ?

> 1)  mpirun -np 2 ./john -format=raw-md5 -i:alpha10 -pot=2millions.pot 
> 2millions.txt
>   -> crack 5665 hashes
> 
>   (After this test 2millions.pot was deleted of course)
> 
> 2) mpirun -np 4 ./john -format=raw-md5 -i:alpha10 -pot=2millions.pot 
> 2millions.txt
>  -> crack 6591 hashes
> 
>  (After this test 2millions.pot was deleted of course)
> 
> 3) mpirun -np 8 ./john -format=raw-md5 -i:alpha10 -pot=2millions.pot 
> 2millions.txt
>  -> crack 7223 hashes
> 
> 
>  Why I get only 1558 more hashes cracked with 8 cores than 2 cores ???

If you think about it, what you did is EQUAL TO the following in a 
non-mpi environment:

1. run plain non-mpi john for 10 minutes. It will crack about 5665 
hashes in 10 minutes.
2. *restore* the job and let it run 10 minutes more. It will crack some 
more hashes, totalling about 6591 hashes in a total of 20 minutes.
3. *restore* the job and let it run 20 minutes more. It will crack some 
more hashes, totalling about 7223 in a total of 40 minutes.

We did the exact same thing, right? But it took 40 minutes.

>  Is it possible to optimise MPIrun to work better with 4 cores and much 
> much better with 8 cores ? 
>  Otherwise, why use 8 cores if I get only few hashes cracked more ?

It already does. You did the same job in five minutes instead of 40!

cheers