Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <af8e84072e19353bafea50ca9539523b@smtp.hushmail.com>
Date: Wed, 08 Jul 2015 13:40:59 +0200
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: PHC: Lyra2 on GPU

On 2015-07-08 11:23, Agnieszka Bielec wrote:
> 2015-07-08 4:15 GMT+02:00 magnum <john.magnum@...hmail.com>:
>> On Mon, Jul 06, 2015 at 01:15:54AM +0200, magnum wrote:
>>> A test run would be effectively the same. A crack run would become:
>>> reset(db)
>>> self-tests
>>> reset(db)
>>> crack mode
>>
>> This is implemented now (without the "int self_test") in 38af431 (formats in
>> e19d87e).
>>
>> Agnieszka, you will likely have to adopt your code a little for it to work
>> at all.
>
> I'm not sure if I understand correctly. now, in the first reset() I
> have to autotune not for self-tests but  only for cracking mode and
> not autotune in the second reset()?

Exactly, except s/have to/can/. But I'd strongly recommend it.

db can still be NULL and if it is, you know this is a benchmark run. In 
that case you could (should) autotune for that.

Solar, Sayantan,

I'm contemplating another change: For benchmark, we can build a little 
fake db that contains whatever test-vectors will be used (and obviously 
pass that). This will simplify the formats' code not only for Agnieszka 
but also for eg. Sayantan's GPU-side mask/compare formats. There would 
be no need for benchmark kludges in the formats. This would also support 
things like "--test --cost=xx" with no special code in the formats.

magnum

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.