Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Sat, 14 May 2011 12:34:36 -0500
From: "JFoug" <>
To: <>
Subject: Re: NT_fmt bug

I hope to get that update out today.  I have MD5/MD4/SHA1 intrinsic code 
incorporated.  There are also numerous other changes/improvements.

I have problems with md5_gen(12), (13) and all of the SHA when building 
intrisic on 64 bit (but it works fine for 32 bit intrisic builds, ???)  At 
this time, I simply have commented out the 12/13 and 21-26 md5_gen if built 
with 64 bit intrinsics.

However, every other test is 100% fully working.  The intrisics are fully 
integrated into all build formats.  At this time, ANY 64 bit build on an x86 
uses the intrisics.  Then on 32 bit builds, there is a choice, and there are 
2 x86-sse.h files (x86-sse.h and x86-ssei.h).    Within the x86-64.h there 
are a couple of build options, depending upon if built using gcc or if built 
using some other compiler.  The main difference right now, is that non-gcc 
builds will have MD4_para and MD5_para set to 3.  All other builds set para 
up to 2.

I still have some work to do to get the diffs right, but then I should put 
that up on the wiki this afternoon (I hope).

There have been many changes to Makefile (and some other config).  Some of 
this is done by best guess on my part. Likely, there will be adjustments 
which will have to be made.


From: "magnum"  Sent: Saturday, May 14, 2011 10:32 AM
>Great, thanks!
>I'm just waiting for JimF to complete some good work now and
>then, in one shot or a couple, we'll get a Jumbo out that will really
>rock. I have code that drops the saved_plain for both ISO8859-1
>and UTF-8 and this makes up for some serious boost!

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.