[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Thu, 5 Mar 2015 19:47:40 +0300
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Xeon Phi
Hi,
I've finally cleaned up and committed my Xeon Phi support patch for
descrypt into the core tree, along with addition of the linux-mic make
target.
The introduction of extra intrinsics is straightforward:
http://cvsweb.openwall.com/cgi/cvsweb.cgi/Owl/packages/john/john/src/DES_bs_b.c.diff?r1=1.42;r2=1.43
The speed is not so great, but we knew this before:
[user@...er-mic0 user]$ LD_LIBRARY_PATH=. ./john -te=1 -form=descrypt
Will run 240 OpenMP threads
Benchmarking: descrypt, traditional crypt(3) [512/512]... DONE
Many salts: 80170K c/s real, 333233 c/s virtual
Only one salt: 7660K c/s real, 61684 c/s virtual
(This is on our 5110P.)
The icc-generated assembly code for DES_bs_b.c looks sane to me (about
as good as what we're used to seeing for AVX, or maybe even better since
there are 32 ZMM registers). I also experimented with adjusting
DES_bs_cpt in DES_bs.h, and the previous default of 32 turned out to be
nearly optimal so I left it intact.
magnum - please merge these changes into jumbo. Thanks in advance.
This will provide a baseline for addition of MIC intrinsics to other
parts of John the Ripper jumbo tree, which is a task we have listed
among our ideas for this year's GSoC projects:
http://openwall.info/wiki/ideas#John-the-Ripper-support-for-Xeon-Phi
... and perhaps this will need to be merged with another task, if done
under GSoC, since at this time it feels like too little work for GSoC,
or maybe the scope should be expanded in some other way - suggestions
are welcome.
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.
