lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  PHC 
Open Source and information security mailing list archives
 
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 23 Apr 2015 12:55:16 +0200
From: Milan Broz <gmazyland@...il.com>
To: discussions@...sword-hashing.net
Subject: Re: [PHC] Re: Updated tests document (version 2)

On 04/23/2015 11:48 AM, Solar Designer wrote:
> On Thu, Apr 23, 2015 at 10:41:51AM +0200, Milan Broz wrote:
>> I updated PHC candidates test with latest run
>>
>>  https://github.com/mbroz/PHCtest/blob/master/output/phc_round2.pdf
>>
>> - many code updates announced here (including Argon, Catena, MAKWA, etc)
>> - added parallel test
> 
> This includes PPC now.  I think it should go up to a much higher number
> of threads on the PPC machine - whatever it supports in hardware (64?)

Actually I run this (I used up to 128 CPUs of 160) but lost output somewhere by mistake.
This run takes a long time and I need to synchronize it with other users.
But it scaled pretty well. (Will try to rerun this.)

Moreover PPC is quite special here (thread per core is configurable,
here it is 8). It would be nice to run this on x86_64, but these
systems are quite busy.

Maybe it would be nice project for some student (we have access
to quite big systems through university and Red Hat has also some big
machines internally) but that will not happen in short-term.
Maybe for testing optimizations for PHC finalist code?
(I would like to have portable test code, at least for Windows and Linux.)

(And I already mentioned I am also interested in lower end of the
spectrum - small systems, including ARM, ARM64.
Neon optimization for Raspberry2 would be nice as well.)

>> - I used newer machine for test (Lenovo x240 i7 CPU)
> 
> This is AVX2 capable, right?  But you're not building for AVX2 yet?

yes, it is AVX2 capable, but -mnative seems to enable only -mavx.

> (And this reminds me that I should submit AVX2 code for yescrypt.)
> 
>> - Added pwxround=2 yescrypt to normalized test
> 
> I think it'll be relevant to all other tests as well. ;-)  And even
> PWXrounds=1 too. ;-)

But then I should test other algorithms with BLAKE2-full,
try BlaMka and many other tweaks... :)

(I would like to -  but I have really no time for it now.)

>> - yescrypt now uses -opt version (not reference)
>> (I did not add it to other test these were meant to use defaults).
> 
> I don't understand.  What "other test" did you not add this to?

That is wrong comment, please ignore it.
This was meant for adding yescrypt with reduced rounds, not to yescrypt-opt.
(Obviously I replaced -ref with -opt everywhere).

Thanks,
Milan

Powered by blists - more mailing lists