john-users - Openwall/Parallella Google Summer of Code project opportunity

Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20130428011019.GA15090@openwall.com>
Date: Sun, 28 Apr 2013 05:10:19 +0400
From: Solar Designer <solar@...nwall.com>
To: announce@...ts.openwall.com
Cc: john-users@...ts.openwall.com
Subject: Openwall/Parallella Google Summer of Code project opportunity

Hi,

The Parallella project has kindly agreed to provide co-mentorship and
hardware for a Google Summer of Code project under Openwall involving
development for the brand new Parallella board.

You might have heard of Parallella from their recent Kickstarter
campaign, where they have successfully raised funds to develop
power-efficient parallel computing boards with their own 16-core and
64-core Epiphany chips as well as Xilinx Zynq FPGAs.  (The boards are
just starting to ship now.)

The Epiphany chip uses a novel architecture: each of its 32-bit RISC
cores has 32 KB of local memory, yet each core can also access any other
core's local memory.  The architecture is designed to scale to up to
4096 cores per chip.  The emphasis was on reaching high floating-point
performance (single-precision multiply-add) per Watt.  In the GSoC
project proposed under Openwall, we'll focus on cryptographic uses
instead (making good use of the local memory, but unfortunately not of
the FPUs).  Success of this project and any shortcomings we identify
might help shape up future revisions of the Epiphany architecture.
(In fact, I've already posted to the Parallella forum some crazy ideas
on how the Epiphany ISA may be made more suitable for symmetric crypto.)

We're looking to implement bcrypt and scrypt cracking and Litecoin
mining on Epiphany.  For scrypt (and Litecoin), we'll need to take
advantage of its deliberate time-memory tradeoff.  With it, the 32 KB
of local memory is sufficient for Litecoin mining (with only a ~2x
performance hit), despite of Litecoin's use of scrypt at 128 KB (when
implemented on regular x86 CPUs).

Here are the specifics (and some additional ideas):

http://openwall.info/wiki/ideas#Support-for-coprocessor-boards-for-John-the-Ripper

Ideally, all of this should be integrated with John the Ripper, which
may run either on the ARM CPU on the Parallella board and access the
Epiphany chip as a coprocessor, or it may run on a traditional desktop
computer and access the entire Parallella board as a coprocessor.
(The boards readily have Linux running on the ARM, so putting JtR in
there shouldn't be difficult.)

Here's the announcement of GNU Radio/Parallella and Openwall/Parallella
GSoC project opportunities from Parallella:

http://www.parallella.org/2013/04/26/google-summer-of-code-opportunities/

Also see the Parallella Community forums:

http://forums.parallella.org

and website of the company behind this project, Adapteva:

http://www.adapteva.com

Student applications are accepted by May 3rd 19:00 UTC from the Google
Summer of Code website:

http://www.google-melange.com/gsoc/org/google/gsoc2013/openwall

Please don't wait until the last moment - apply sooner, then edit your
application before the deadline if necessary.  Also, besides applying
formally, please contact us to discuss the specifics of your proposal
and arrange for a qualification task (to be completed before the student
selection deadline of around May 20).

Alexander
Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.