Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 19 Sep 2005 04:09:48 +0400
From: Solar Designer <solar@...nwall.com>
To: john-users@...ts.openwall.com
Subject: Re: Suggestions for XEON performance?

On Sat, Sep 17, 2005 at 04:44:59PM -0400, Douglas F. Calvert wrote:
>  I am a little surprised with the performance of john on the XEON
> (proc/cpuinfo is attached at the bottom) and linux. I guess I expected
> the performance to be much better on the XEON than it is. Maybe I am
> doing something wrong with my compilation.

Unfortunately, you haven't mentioned what kind of performance you're
getting and what you expected.

P4 Xeons should give almost the same performance as non-Xeon P4s.

P4s in general are slower than P2/P3 per-MHz.  (The only big exception to
this is P4's much higher memory bandwidth, but it is irrelevant for John.)

> Does anyone have any
> reccomendations for which target to build? I have been using
> linux-x86-mmx-elf.

That's correct.

(I am planning to add "linux-x86-sse-elf" which would be the right
choice for P4s.  With that, they would become only a little bit slower
than P3s per-MHz.  Stay tuned.)

> The gentoo safe cflags wiki lists a few
> optimizations for the gcc on the XEON but they made it a little
> slower:
> 
> CFLAGS="-march=nocona -O2 -fomit-frame-pointer -mmmx -msse2 -mfpmath=sse -pipe"

C compiler flags are largely irrelevant when most of the performance
critical code is written in assembly right away.

> Were my expectations too high for the XEON?

I have no idea what your expectations were.

> Also should I use gcc 3.4 or 4.0?

It shouldn't matter.

Although if you're really concerned, gcc 2.95.3 is known to generate
slightly better code than gcc 3.x (haven't tried 4.0) for LM hash key
setup.  This may make a 10% difference or so at LM hashes.  Other hash
types should remain unaffected.

> john is works wonderfully and is definitely fast enough I was
> just under the impression that I was going to see a bigger speed
> boost...

I'm not sure what you're comparing the Xeons against.  If you're
comparing them against HT-capable non-Xeon P4s, there should be no
difference.

Your "cpuinfo" suggests that you're on a Dual-Xeon system with 4 logical
CPUs.  You do realize that a single instance of John will use just one
of the logical CPUs, correct?  You may get a 2.5x better overall
performance by starting 4 instances of John.  Please refer to the list
archives for a discussion on how the task may be split for 4 CPUs.  The
messages subject was "trivial parallel processing (4 CPUs)".

-- 
Alexander Peslyak <solar at openwall.com>
GPG key ID: B35D3598  fp: 6429 0D7E F130 C13E C929  6447 73C3 A290 B35D 3598
http://www.openwall.com - bringing security into open computing environments

Was I helpful?  Please give your feedback here: http://rate.affero.net/solar

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.