Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20130407151243.GA30817@openwall.com>
Date: Sun, 7 Apr 2013 19:12:43 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: Fix for NT format (SSE builds, 32 bit)

On Sun, Apr 07, 2013 at 07:09:38PM +0400, Solar Designer wrote:
> On Sun, Apr 07, 2013 at 11:03:36AM -0400, jfoug@....net wrote:
> > Here is a fix to the bleeding tree, for the NT format.  It has been failing in crypt_all
> 
> Thanks!  I think this code will run slightly faster if we correct the
> offset rather than move the mov instruction.

Oops, forget that.  Obviously, the wrong offset was not the only bug.
"popa" also trashes the just loaded %eax, so we can't easily have that
instruction at that place - or we need to replace the pusha/popa with
individual push/pop instructions, which may well be faster too.
Luckily, this code is not mine, so I am not the one responsible, nor the
one to try improving it. ;-)

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.