Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Tue, 10 Sep 2013 04:59:15 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: Rafael's weekly report #13

Rafael,

On Tue, Sep 10, 2013 at 02:06:07AM +0200, Rafael Waldo Delgado Doblas wrote:
>  Priorities:
> 1. Improve salsa20_8 ASM.

I hope by "improving" it you mean starting to use macros to the greatest
extent possible, committing that, and then starting with instruction
scheduling (and consequently with making macros larger, so that you can
schedule instructions inside them as required to avoid stalls), and
committing that too.  I'd like to see at least two commits along the way
(first with maximum use of macros, then with instruction scheduling).

> 2. Find the way to compile ARM and Epiphany codes with automake and only
> one call to make.

As described above, these two tasks shouldn't take you more than two
days, combined.  Surely good asm code may take many days to polish, but
in a day or two you ought to be able to bring it to a state that at
least looks similar (in terms of approach taken) to Katja's bcrypt code.
So please try, and please report on that on Wednesday or Thursday, so
that we can move forward with further aspects of the task.

Also, what has happened to my request to keep the pure C version of the
code in there and usable (as a compile-time option)?  Did you read that
message at all?  (This is a reason why I prefer that you do inline
quoting - that way, I am more confident that you have read the message
that you reply to.)

BTW, you still have that byte-by-byte blkcpy() - and no plans on fixing
that?  It's a trivial and quick task (one hour?), yet something you
could do as well.  Luckily, in your current code blkcpy() is only used
in SHA-256 and PBKDF2, not in SMix, yet you could make this trivial
optimization to it.

I understand that you might be frustrated by how we've been using your
time this summer for not-so-important tasks only, yet this is largely a
result of my unhappiness with the slow progress and not so great quality
of your work.  I simply do not dare to throw more tasks at you yet, and
GSoC is already almost over.

For example, there's an easy optimization task for JtR jumbo code that
runs on all platforms (so not Parallella specific).  It'd be a one
evening task for me.  If you feel bored with the current stuff and need
to do something immediately useful, to boost your confidence and thus
further work (including this Litecoin mining on Parallella stuff), let
me know and I can describe this task to you.  It might not be a one
evening task for you, but it shouldn't take you more than a couple of
days either.

So let me know.

Thanks,

Alexander

Powered by blists - more mailing lists

Your e-mail address:

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