Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 25 Apr 2013 20:49:13 +0200
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: ICC performance regression

I rebooted to Ubuntu and tested a 32-bit build. Works like a champ, OMP or not. It's even faster than before the intrinsics changes. You must have something messed up in your tree.

magnum


On 25 Apr, 2013, at 20:20 , magnum <john.magnum@...hmail.com> wrote:

> $ git grep SSESHA1body *.S
> sse-intrinsics-32.S:#define SSESHA1body     _SSESHA1body
> sse-intrinsics-32.S:# -- Begin  SSESHA1body
> sse-intrinsics-32.S:    .globl SSESHA1body
> sse-intrinsics-32.S:SSESHA1body:
> 
> Well there is something, but I can't test it here. Please elaborate. You should be able to build the same file using the sse-intrinsics-32.S make target, and perhaps check if the perl script does something wrong? Although from git history it looks like the exact same script was used last time.
> 
> magnum
> 
> 
> On 25 Apr, 2013, at 19:52 , "jfoug" <jfoug@....net> wrote:
> 
>> SSESHA1body is no longer appearing in the .S file ???  Whats up with that?
>> 
>> -----Original Message-----
>> From: magnum [mailto:john.magnum@...hmail.com] 
>> Sent: Thursday, April 25, 2013 2:10
>> To: john-dev@...ts.openwall.com
>> Subject: Re: [john-dev] ICC performance regression
>> 
>> On 25 Apr, 2013, at 1:30 , Solar Designer <solar@...nwall.com> wrote:
>>> On Thu, Apr 25, 2013 at 01:12:19AM +0200, magnum wrote:
>>>> Old pre-built files, icc 12.1.4:
>>> [...]
>>>> Benchmarking: FreeBSD MD5 [128/128 SSE2 intrinsics 12x]... DONE
>>>> Raw:	39204 c/s real, 39204 c/s virtual
>>> [...]
>>>> gcc 4.7.2, -native target:
>>> [...]
>>>> Benchmarking: crypt-MD5 [128/128 AVX intrinsics 12x]... DONE
>>>> Raw:	36936 c/s real, 36936 c/s virtual
>>> 
>>> This is pretty significant difference in favor of old icc, and not all 
>>> CPUs have AVX, so I think we should simply continue to use old icc to 
>>> prebuild the files.
>> 
>> This requires someone having an older version. I haven't found one yet.
>> 
>> Until now we have compared icc using -O3 (25 *minutes* compile time per
>> file), to gcc using just -O2 (compiling in 3 seconds). I will try some
>> different versions of icc as well as MD5_PARA values (very time consuming),
>> but also different sets of options to gcc and see where we end up.
>> 
>>> Why the name discrepancy, though?  There was no intent to rename this 
>>> format to crypt-MD5, was there?  If I rename it, I'll use md5crypt, 
>>> including in the printed name.
>> 
>> I changed label now but not the name so it's just cosmetical. I found it
>> misleading when loading an AIX {smd5} hash and it answered
>> 
>> Loaded 1 password hash (FreeBSD MD5 [128/128 AVX intrinsics 12x])
>> 
>> ...as if that was specifically what's been loaded. We can use md5crypt
>> instead though.
>> 
>> magnum
>> 
>> 
>> 
> 
> 
> 


Powered by blists - more mailing lists

Your e-mail address:

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