[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20130730012951.GA14512@openwall.com>
Date: Tue, 30 Jul 2013 05:29:51 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: Parallella: bcrypt
Katja -
On Tue, Jul 30, 2013 at 12:44:09AM +0200, Katja Malvoni wrote:
> Speed is 1175 c/s.
If you can do 1210 c/s while also using the two ARM cores for another
160 c/s (normally, they do 166 c/s in an OpenMP-enabled build of JtR),
this gives us a nice 5000 c/s on 64-core. ;-)
1210*4 + 160 = 5000
That said, there is indeed more potential in using the FPGA than in
further Epiphany and ARM code optimizations, so you should in fact focus
on the FPGA soon. Here's some code from our 2011 project:
http://openwall.info/wiki/crypt-dev/files
This includes Eksblowfish in Verilog using BlockRAM in Spartan-6 or
Virtex-6, but otherwise mostly unoptimized. Here are some thoughts on
optimizing it:
http://www.openwall.com/lists/crypt-dev/2011/08/21/1
You should also cleanup your bcrypt on Epiphany code and submit it for
inclusion in the main bleeding-jumbo branch. BTW, I think the make
target should be simply "linux-parallella" rather than the lengthy
"linux-arm32le-neon-parallella".
Thanks,
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.
