john-dev - Re: [GSoC] John the Ripper support for PHC finalists

Follow @Openwall on Twitter for new release announcements and other news

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAKGDhHU5k-ff_O5xUmz7+Pr5RNyi_v8BhJqFWvOXy4H7CF2qdA@mail.gmail.com>
Date: Sat, 18 Apr 2015 23:59:58 +0200
From: Agnieszka Bielec <bielecagnieszka8@...il.com>
To: john-dev@...ts.openwall.com
Subject: Re: [GSoC] John the Ripper support for PHC finalists

2015-04-18 20:45 GMT+02:00 Agnieszka Bielec <bielecagnieszka8@...il.com>:
> If benchmark encrypts one password on various gws, bencharking will
> give better result than it should be due to less random memory access
> I'm trying to investigate this

It's true. All the time I was receiving wrong benchmarking results
because if we have the same password to encrypt in various GPU's units
function H which randomly accesses the memory works faster due better
coalescing
I made an experiment.
previously I wrote:

> if in fmt_test are 2 records, it is fast
> when I add 10 records with the same cost, it slows down

If I remove function H from the kernel all the time the speed is the same

I've modified the file bench.c to random all keys and I think I've
received  true results now

[a@...er run]$ ./john --test --format=pomelo-opencl --dev=1 --cost=2:2,2:2
Benchmarking: pomelo-opencl, POMELO [POMELO OpenCL (inefficient,
development use only)]... max cost 2
Device 1: Tahiti [AMD Radeon HD 7900 Series]
Local worksize (LWS) 64, global worksize (GWS) 4096
DONE
Speed for cost 1 (N) of 2, cost 2 (r) of 2
Raw:    80313 c/s real, 8192K c/s virtual

[a@...er run]$ ./john --test --format=pomelo-opencl --dev=5 --cost=2:2,2:2
Benchmarking: pomelo-opencl, POMELO [POMELO OpenCL (inefficient,
development use only)]... max cost 2
Device 5: GeForce GTX TITAN
Local worksize (LWS) 64, global worksize (GWS) 8192
DONE
Speed for cost 1 (N) of 2, cost 2 (r) of 2
Raw:    82671 c/s real, 82671 c/s virtual

not as bad ( comparing to previous ) in my opinion

I think that there are 3 possible solutions
1). modify bench.c like I've modified (but, are there a hashes which
cannot receive some specific password values? )
2). spefify more tests cases for fmt_tests (but, sometimes we will
need to pass 100 tests cases and sometimes more)
3). make fmt_main to generate password values if the answer in question is yes

what are your opinions?

BTW I think that I should make some charts comparing the speed of GPU
with CPU in the future

thanks

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.