Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sat, 23 Jun 2012 23:51:57 +0200
From: Frank Dittrich <>
Subject: Re: Some relbench results for linux-x86-sse2 (jumbo5 vs.
 upcoming jumbo6)

On 06/23/2012 11:37 PM, Solar Designer wrote:
> Frank,
> On Sat, Jun 23, 2012 at 11:11:09PM +0200, Frank Dittrich wrote:
>> And these are the worst regressions:
>> $ ./relbench -v j5-sse2-mod.txt j6-sse2.txt|grep "^Ratio:"|cut -f
>> 2-|sort -n|head -n 10
> ...
>> 0.61111 real, 0.61111 virtual	MediaWiki md5($s.'-'.md5($p)):Many salts
>> 0.87012 real, 0.87891 virtual	phpass MD5 ($P$9):Raw
> ...
>> 0.88684 real, 0.89589 virtual	MySQL:Raw
> ...
>> 0.90846 real, 0.90857 virtual	dynamic_27: FreeBSD MD5:Raw
>> 0.91026 real, 0.91026 virtual	Generic salted MD4:Many salts
>> 0.92756 real, 0.92756 virtual	dynamic_0: md5($p) (raw-md5):Raw
> Can you post the actual output for these (for both versions)?

I'll attach 3 files with benchmark output, the j6 one (current git),and
two j5 (original and modified by benchmark-unify).
UPDATE: I had to compress them to educe the size.

> Are these regressions fully reproducible if you re-run both versions?

I just tried the first two, they are fully reproducible.
At least for MediaWiki, I suspect JimF's patches to support
--regen-lost-salts, but I may be wrong.

> What make target did you use?  linux-x86-sse2 or linux-x86-sse2i?

sse2, without OMP.
I intended to add more benchmarks, but I thought I better try this with
x86-64 on my Laptop with an i7-2820QM CPU @ 2.30GHz.


Download attachment "benchmarks-x86-sse.tgz" of type "application/x-gzip" (7667 bytes)

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.