Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Thu, 25 Jun 2015 08:57:42 +0300
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: optimizing bcrypt cracking on x86

On Wed, Jun 24, 2015 at 07:50:15PM -0400, Alain Espinosa wrote:
> Yes. Attached are all my CPUs benchmarks. There are 3 different architectures: core 2 duo, Sandy Bridge and Haswell. Can you make direct comparison between core i5 and core i7? Perhaps my assembly code is suboptimal.

Thanks.  I didn't mean to say your code is worse than mine.  It probably
is OK, possibly even better than mine - for your CPU at least.

> One thing I notice is that you first try 32bits asm code. I use 64bits asm code only. Can this make a difference?

Sure.  When I first ported the 32-bit mode MMX2 code to 64-bit mode, it
actually ran slower than it did in 32-bit mode on the same Haswell CPU -
there were extra prefixes.  I then revised it slightly to avoid the
prefixes, and it became same speed.  And then I could proceed to make
use of the extra 8 GPRs and of BMI, which I did - but with limited
success, as I documented.

I merely wanted to share the experiments and findings and ideas so far.
I do not claim to have achieved much.

In a way, I am relieved that there doesn't appear to be much room for
further attack speedup on current x86 CPUs.

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.