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 16:56:08 +0800
From: Lei Zhang <>
Subject: Re: JtR on Power

> On Jul 9, 2015, at 12:00 PM, Lei Zhang <> wrote:
> Strangely, MD5, SHA256 and SHA512 become even slower. Currently there're some inefficient emulation for several intrinsics, notably int64 shift, which very likely slows down SHA512 here. I try to do some optimization and see how it goes.

With some optimization, the result gets better. 

OpenMP disabled this time:

Benchmarking: Raw-MD4 [MD4 32/64]... DONE
Raw:	3552K c/s real, 3588K c/s virtual

Benchmarking: Raw-MD5 [MD5 32/64]... DONE
Raw:	2784K c/s real, 2812K c/s virtual

Benchmarking: Raw-SHA1 [SHA1 32/64]... DONE
Raw:	3661K c/s real, 3661K c/s virtual

Benchmarking: Raw-SHA256 [SHA256 32/64 OpenSSL]... DONE
Raw:	2306K c/s real, 2330K c/s virtual

Benchmarking: Raw-SHA512 [SHA512 32/64 OpenSSL]... DONE
Raw:	1868K c/s real, 1887K c/s virtual

Benchmarking: Raw-MD4 [MD4 128/128 AltiVec 4x]... DONE
Raw:	10755K c/s real, 10649K c/s virtual

Benchmarking: Raw-MD5 [MD5 128/128 AltiVec 4x]... DONE
Raw:	9073K c/s real, 9258K c/s virtual

Benchmarking: Raw-SHA1 [SHA1 128/128 AltiVec 4x]... DONE
Raw:	7964K c/s real, 8127K c/s virtual

Benchmarking: Raw-SHA256 [SHA256 128/128 AltiVec 4x]... DONE
Raw:	5849K c/s real, 5908K c/s virtual

Benchmarking: Raw-SHA512 [SHA512 128/128 AltiVec 2x]... DONE
Raw:	2711K c/s real, 2711K c/s virtual

Speedup of each format:
MD4	3.0
MD5	3.3
SHA1	2.2
SHA256	2.5
SHA512	1.4

I'll take a closer look at AltiVec's manual and see if I can squeeze a little more performance out of it.


Powered by blists - more mailing lists

Your e-mail address:

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