Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CA+TsHUA7wm1AmZkFLJFtj0dNUfV=_VQ7hF6E=iGcTbiUA-QVEA@mail.gmail.com>
Date: Mon, 4 Jun 2012 21:48:12 +0530
From: SAYANTAN DATTA <std2048@...il.com>
To: john-users@...ts.openwall.com, NeonFlash <psykosonik_frequenz@...oo.com>
Subject: Re: JtR OMP Questions

On Mon, Jun 4, 2012 at 7:32 PM, NeonFlash <psykosonik_frequenz@...oo.com>wrote:

> I have a few questions regarding, JtR's OMP Feature:
>
> Is there a way to control how many logical cores of the multi core CPU are
> used by john-omp.exe? By default, it uses all the available logical cores
> of the CPU but I would like to control this manually.
>
>
> I did not see any such option listed in the help.
>
> Does the OMP feature apply only to few hash types? For instance, I can see
> the multiple CPU Cores being used for few hash types like DES, Blowfish but
> not for others like dynamic_12 and so on. An example from the benchmark:
>
> C:\JTR\run>john-omp.exe --format=DES --test=3
> Benchmarking: Traditional DES [128/128 BS SSE2]... (8xOMP) DONE
> Many salts:     8770K c/s real, 1162K c/s virtual
> Only one salt:  7713K c/s real, 1105K c/s virtual
>
>
> C:\JTR\run>john-omp.exe --format=dynamic_12 --test=3
> Benchmarking: dynamic_12: md5(md5($s).md5($p))  (IPB)  [SSE2i 10x4x3]...
> DONE
> Many salts:     2723K c/s real, 2717K c/s virtual
> Only one salt:  1966K c/s real, 1969K c/s virtual
>
> Is there a way to use multiple CPU Cores with a hash type such as
> dynamic_12?
>
>
> What is the difference between OMP and MPI?
>
> On linux platform, I noticed in the Makefile of JtR, that we can enable
> the OMPFLAGS to compile the JtR with OpenMP Compatibility and there is also
> an option to use MPI with OpenMP.
>
> However, I would like to understand the difference between the two?
>
> Thanks.


Set OMP_NUM_THREADS=1,2,3,4....

eg.
bash$ OMP_NUM_THREAD=3 ./john -te -fo=raw-md5
it will use only 3 threads.

While openMP is generally used for fine grain parallelism like within a
specific format. On the other hand MPI is generally used for task level
parallelism e.g two or more instances of john.

Regards,
Sayantan

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.