phc-discussions - Re: [PHC] Comparing speed of entries

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 linux-cve-announce 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]

Message-ID: <54F47F64.2000406@larc.usp.br>
Date: Mon, 02 Mar 2015 12:19:00 -0300
From: Marcos Simplicio <mjunior@...c.usp.br>
To: discussions@...sword-hashing.net
Subject: Re: [PHC] Comparing speed of entries

On 02-Mar-15 02:15, Ben Harris wrote:
> On 2 March 2015 at 12:30, Bill Cox <waywardgeek@...il.com> wrote:
> 
>>
>> There seems to be some confusion about how to compare the speed of entries
>> in the PHC.  IMO, we should:
>>
>> - Compare entries when using the _same_ hash function, the same number of
>> rounds, and the same number of threads.
>>
>> With this simple rule, it is easy to see that Yescrypt is the fastest,
>> Lyra2 is the second fastest.
>>
>>
> The comparisons are more difficult for the entries that aren't 'sequential
> memory hard', e.g. Catena's lambda variable. I'd like to see a
> normalisation of these entries to be the number of invocations of the hash
> primitive. You then have two independent variables 'memory' and 'hash
> invocations' and the dependent variable is 'time' or 'bandwidth'.
> 
> Also looking at the numbers for the <= L1 and <= L2 memory sizes (for the
> benchmarking machine).

I do agree that some kind of normalisation is necessary for a fair
comparison.

We tried to consider the "number of invocations of the hash primitive"
in our benchmarks of latest reference guide for Lyra2, so maybe Table 3
(page 57) may help on this task. The graphs are not normalized to the
same hash function/number of rounds, though. I feel, that doing so may
be reasonable when the scheme uses a "generic" hash function, but that
may not be the case when one of the novelties of the proposal is the
underlying hash itself...

BR,

Marcos.