Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 15 Jun 2012 20:02:42 +0200
From: Tavis Ormandy <taviso@...xchg8b.com>
To: john-dev@...ts.openwall.com
Subject: Re: Re: [patch] optional new raw sha1 implemetation

"jfoug" <jfoug@....net> wrote:

> On Friday, June 15, 2012 11:45 AM Tavis Ormandy wrote
> > On Fri, Jun 15, 2012 at 04:54:15PM +0200, Tavis Ormandy wrote: ... I
> > made some slight adjustments to the scheduling, please use the attached
> > version instead.
> 
> Travis,
> 
> Please change this:
> 
> - #if defined(__SSE4_1__) && defined(__GNUC__) && defined(__linux__) + #if
> defined(__SSE4_1__) && defined(__GNUC__) && (defined(__linux__) || defined
> (__CYGWIN32__))
> 
> I have built on cygwin (with -msse4 instead of -msse2), and it builds just
> fine. It is not faster (about 5% slower), but builds fine, and tests fine.
> 
> Also, this should be changed.  This is the proper benchmark setting for a
> non-salted format:
> 
> -        .benchmark_length   = 1, +        .benchmark_length   = -1,
> 
> With that benchmark testing changed, my cygwin build now benches this new
> code about 5% faster than the raw-sha1 from bleeding.
> 

Ahh, I guess I didn't understand that parameter. Thanks!

5% improvement is nice, but I'm dissapointed it's not higher. Could you send
me a .o file? I would like to disassemble it and see where GCC goes wrong,
hopefully there's a simple way to fix it.

Version 3 attached.

Tavis.

-- 
-------------------------------------
taviso@...xchg8b.com | pgp encrypted mail preferred
-------------------------------------------------------


View attachment "taviso_fmt_plug.c" of type "text/x-csrc" (23620 bytes)

Powered by blists - more mailing lists

Your e-mail address:

Powered by Openwall GNU/*/Linux - Powered by OpenVZ