Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Tue, 24 Jul 2012 14:55:43 +0200
From: Lukas Odzioba <lukas.odzioba@...il.com>
To: john-dev@...ts.openwall.com
Subject: Re: Current -fixes GPU formats vs TS

2012/7/23 magnum <john.magnum@...hmail.com>:
> Here are current figures from 1.7.9-jumbo-6-fixes branch, using GTX 570
> on bull (those that pass was snipped):
>
> form=mscash-cuda                  guesses: 1216 time: 0:00:00:00 :
> Expected count(s) (1430)  [!!!FAILED!!!]
> form=mscash2-cuda                 guesses: 1050 time: 0:00:00:08 :
> Expected count(s) (1479)  [!!!FAILED!!!]
> form=raw-sha224-cuda              guesses: 1476 time: 0:00:00:00 :
> Expected count(s) (1500)  [!!!FAILED!!!]
> form=raw-sha256-cuda              guesses: 1476 time: 0:00:00:00 :
> Expected count(s) (1500)  [!!!FAILED!!!]
> form=raw-sha512-cuda              guesses: 1476 time: 0:00:00:00 :
> Expected count(s) (1050)  [!!!FAILED!!!]
>
> The numbers 1050, 1216 and 1476 rings a bell, I believe they correspond
> to certain max. lengths. Maybe the remaining problem is just in what TS
> believes is the maximum length?
>
>
> form=pwsafe-cuda                  guesses:    0 unk unk : Expected
> count(s) (1485)  [!!!FAILED!!!]
> form=md5crypt-opencl              guesses:    0 unk unk : Expected
> count(s) (1500)  [!!!FAILED!!!]
> form=phpass-opencl                guesses:    0 unk unk : Expected
> count(s) (1470)  [!!!FAILED!!!]
> form=nt-opencl                    guesses:    0 unk unk : Expected
> count(s) (1478)  [!!!FAILED!!!]

md5crypt and phpass wasn't changed since jumbo-6 release and they
passed last tests.
This is at least weird for me.
pwsafe-cuda could have some issue, we haven't had tests for it in the
previous TS.

Raw-sha244,256 prevoiusly passed all tests but I limited plaintext
length and their current results seems ok to me.
One of the mscash/mscash2 cuda format had some problems earlier and
there wasn't fixed yet.

Anyway I am unhappy with those results but thanks for testing it!

P.S.
In my opinion if password length is grater than what format can hande
TS must not treat this as failed tests. This leads to annoying false
tests, and if I tommorow change tests to support length up to 128
chars we will get near 100% failed tests which is not what  expect.


Lukas

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.