Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20140325220928.GA12712@openwall.com>
Date: Wed, 26 Mar 2014 02:09:28 +0400
From: Solar Designer <solar@...nwall.com>
To: john-users@...ts.openwall.com
Subject: Re: john-1.8.0-Win-32 - john.omp.exe - cores used

On Tue, Mar 25, 2014 at 10:43:12PM +0100, -.-PhanTom-.- wrote:
> In /run folder I can invoke ./john fine for the 1.8.0.2-bleeding so it 
> looks like it compiled... but I don't understand, why no dll files at 
> all are in the same dir.

They're not required to be in the same dir.  We're merely adding them
into the same dir in the builds we distribute, because people might not
have Cygwin installed or it might not be a compatible version (and also
because our cygwin1.dll is patched, but that's a separate story).

When you do have Cygwin installed and are building programs, you're
indeed _not_ required to copy Cygwin DLL's for each and every program.
That would largely defeat the purpose of using dynamic linking.

> So I cannot run this in a windows command 
> prompt as I still get the message about the missing crypto dll. Also the 
> normal cygwin.dll is also not present.

As Jim explained, setting up environment variables may be part of Cygwin
setup, which you probably haven't completed.

Doesn't Cygwin's "launch bash" desktop icon setup the env vars, though?

> In cygwin bash ( ~/JtR-Bleeding/run), I get:
> 
> $ ./john --test
> Will run 4 OpenMP threads
> Benchmarking: descrypt, traditional crypt(3) [DES 128/128 SSE2-16]... 
> (4xOMP) DONE
> Many salts:     5262K c/s real, 2245K c/s virtual
> Only one salt:  5080K c/s real, 2364K c/s virtual
> 
> Those test speeds I don't understand at all.....

Yes, it's weird.  I suggest that you debug step by step: try making a
non-OpenMP 64-bit build first and measure its speeds.  Also, look at
speeds you're getting for other hash types, for both builds.

BTW, why don't you enable AVX?  Use an -x86-64-avx make target.

> Good point about the 1 core speed versus 4 cores. I guess i will have to 
> try another performance profile and maybe even overclock a bit.

It's perfectly fine that you should be getting higher clock rate when
using 1 core vs. when using 4.  That's how the CPU is supposed to work.

Alexander

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.