[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20140325211924.GA12404@openwall.com>
Date: Wed, 26 Mar 2014 01:19:24 +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 09:37:09PM +0100, -.-PhanTom-.- wrote:
> testing the mailing list.. hope it works :D
It does.
> I did do that in the win64-build I have been trying to compile myself,
> but with no success yet (see my reply to JimF post on the guide he made
> to compile Bleeding under Cygwin (64bit)
Yes. I've never tried 64-bit Cygwin, but I think the DLL should come
with Cygwin somewhere. Maybe you didn't make a sufficiently complete
install. The bash output showing "-bash: ./john: No such file or
directory" suggests that "john" itself might not be in the current
directory, so I'd double-check that first.
> H:\HA\john179w2\john179\run>john-omp -t
> Benchmarking: Traditional DES [128/128 BS SSE2]... DONE
> Many salts: 11338K c/s real, 3265K c/s virtual
> Only one salt: 10791K c/s real, 3204K c/s virtual
>
> H:\HA\john179w2\john179\run>john -t
> Benchmarking: Traditional DES [128/128 BS SSE2]... DONE
> Many salts: 4506K c/s real, 4635K c/s virtual
> Only one salt: 4312K c/s real, 4425K c/s virtual
>
> So only 152% better with 4 cores for DES in john179w2.
In other words, 2.52x out of 4x max. That's not great, but it's not
that bad either if you recall that we're switching from non-thread-safe
assembly code to thread-safe compiler-generated code on an architecture
with only 8 registers (where compilers have difficulty generating decent
code, especially for something like bitslice DES). A 64-bit build
should provide higher efficiency, since we have 16 registers there.
Also, your CPU probably runs at a higher clock rate with only 1 core in
use (max turbo) than with all 4 in use, so the max speedup is below 4x.
Your bcrypt benchmark suggests that it's about 3.83x, which is rather
good, though.
> H:\HA\john179j5w\john179j5\run>john-omp -test
> Benchmarking: Traditional DES [128/128 BS SSE2]... (4xOMP) DONE
> Many salts: 12315K c/s real, 3276K c/s virtual
> Only one salt: 10764K c/s real, 3087K c/s virtual
>
> H:\HA\john179j5w\john179j5\run>john -test
> Benchmarking: Traditional DES [128/128 BS SSE2]... DONE
> Many salts: 4670K c/s real, 4675K c/s virtual
> Only one salt: 4505K c/s real, 4517K c/s virtual
>
> And around 164% better with 4 cores for john179j5w.
That's 2.64x out of ~3.83x max. I don't know why it performs better
than john179w2 for you. Could be lucky relative placement of its
components in memory.
> I never really got around to learning my way around Linux as on OS, but
> I guess the userfriendlyness has come a long way since I last tried.
> I guess I could try to install Ubuntu in a VmWare sessions and try to
> compile there...
> Would Ubuntu in a WVWare session perform as well as a native Ubuntu
> dualboot installation?
It might perform almost as well.
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.
