Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 10 Jul 2015 17:34:34 +0800
From: Lei Zhang <zhanglei.april@...il.com>
To: john-dev@...ts.openwall.com
Subject: Re: JtR on Power


> On Jul 9, 2015, at 6:59 PM, Solar Designer <solar@...nwall.com> wrote:
> 
> Can you show md5crypt, phpass, sha256crypt, sha512crypt?  And PBKDF2-*?

OpenMP disabled:

[before]

Benchmarking: md5crypt, crypt(3) $1$ [MD5 32/64 X2]... DONE
Raw:	3214 c/s real, 3214 c/s virtual

Benchmarking: phpass ($P$9) [phpass ($P$ or $H$) 32/64 1x2  (MD5_body)]... DONE
Raw:	1438 c/s real, 1438 c/s virtual

Benchmarking: sha256crypt, crypt(3) $5$ (rounds=5000) [SHA256 32/64 OpenSSL]... DONE
Speed for cost 1 (iteration count) of 5000
Raw:	381 c/s real, 381 c/s virtual

Benchmarking: sha512crypt, crypt(3) $6$ (rounds=5000) [SHA512 64/64 OpenSSL]... DONE
Speed for cost 1 (iteration count) of 5000
Raw:	161 c/s real, 161 c/s virtual

Benchmarking: PBKDF2-HMAC-MD5 [PBKDF2-MD5 32/64]... DONE
Speed for cost 1 (iteration count) of 1000
Raw:	1132 c/s real, 1143 c/s virtual

Benchmarking: PBKDF2-HMAC-SHA256 [PBKDF2-SHA256 64/64 OpenSSL]... DONE
Speed for cost 1 (iteration count) of 1000
Raw:	738 c/s real, 738 c/s virtual

Benchmarking: PBKDF2-HMAC-SHA512, GRUB2 / OS X 10.8+ [PBKDF2-SHA512 64/64 OpenSSL]... DONE
Speed for cost 1 (iteration count) of 1000
Raw:	566 c/s real, 566 c/s virtual

[after]

Benchmarking: md5crypt, crypt(3) $1$ [MD5 128/128 AltiVec 4x]... DONE
Raw:	20628 c/s real, 20628 c/s virtual

Benchmarking: phpass ($P$9) [phpass ($P$ or $H$) 128/128 AltiVec 4x1]... DONE
Raw:	10423 c/s real, 10528 c/s virtual

Benchmarking: sha256crypt, crypt(3) $5$ (rounds=5000) [SHA256 128/128 AltiVec 4x]... DONE
Speed for cost 1 (iteration count) of 5000
Raw:	787 c/s real, 787 c/s virtual

Benchmarking: sha512crypt, crypt(3) $6$ (rounds=5000) [SHA512 128/128 AltiVec 2x]... DONE
Speed for cost 1 (iteration count) of 5000
Raw:	553 c/s real, 553 c/s virtual

Benchmarking: PBKDF2-HMAC-MD5 [PBKDF2-MD5 128/128 AltiVec 4x]... DONE
Speed for cost 1 (iteration count) of 1000
Raw:	10864 c/s real, 10864 c/s virtual

Benchmarking: PBKDF2-HMAC-SHA256 [PBKDF2-SHA256 128/128 AltiVec 4x]... DONE
Speed for cost 1 (iteration count) of 1000
Raw:	3576 c/s real, 3576 c/s virtual

Benchmarking: PBKDF2-HMAC-SHA512, GRUB2 / OS X 10.8+ [PBKDF2-SHA512 128/128 AltiVec 2x]... DONE
Speed for cost 1 (iteration count) of 1000
Raw:	1488 c/s real, 1488 c/s virtual


Speedup of each format:

-------------------
md5crypt	6.4
phpass		7.2
sha256crypt	2.1
sha512crypt	3.4
pbkdf2-md5	9.6
pbkdf2-sha256	4.8
pbkdf2-sha512	2.6
-------------------

Well, some of the speedups seem too high. I once doubted if I got other load when benchmarking those non-AltiVec figures, so I retried the 'before' part. But the figures didn't get higher. Perhaps something other than vectorization is playing a role here, but I don't know.


>> Strangely, MD5, SHA256 and SHA512 become even slower.
> 
> I don't know what exact CPUs you're on, but I suspect they are designed
> to run 4 or 8 threads/core (for POWER7 and POWER8, respectively), and
> the impact from not doing so might be more profound for SIMD
> instructions (higher latency) than for scalar ones.

It's Power 8. I disabled OpenMP in this benchmarking.


Lei

Powered by blists - more mailing lists

Your e-mail address:

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