Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 7 Jun 2013 18:39:23 +0200
From: Dániel Bali <balijanosdaniel@...il.com>
To: john-dev@...ts.openwall.com
Subject: Re: sha3-opencl

Hello!

2013/6/6 Lukas Odzioba <lukas.odzioba@...il.com>
>
>
> Since it works I am ok with that speed.
> Just for curiosity, what speed do you get on 7970 ( add -platform=1)
>

Something is very wrong here. It doesn't run on the 7970. Changing
the MAX_KEYS_PER_CRYPT value changes the behavior.
With (1024 * 2048) for MAX_KEYS_PER_CRYPT  it runs on the GeForce card and
then there is a double free error.
With any other value I tested so far (1024 * (1024, 512, 4096)) it results
in a segfault.

With (1024 * 1024) on the 7970 I get a segfault again.
With any value above (1024*1024) for MAX_KEYS_PER_CRYPT  on the 7970 it
results in a failed test:

daniel@...l:~/bleeding-jumbo/JohnTheRipper/src$ ../run/john -test
--format=raw-keccak256-opencl --devices=1
Device 1: Tahiti (AMD Radeon HD 7900 Series)
Local worksize (LWS) 256, global worksize (GWS) 4194304
Benchmarking: raw-keccak256-opencl, Raw Keccak256 [OpenCL (inefficient,
development use only)]... FAILED (get_key(2362204))

I guess something is wrong in the way buffers/arrays are handled, but I'm
not familiar enough with all the parameters to know where the error might
be.


> Please also post a link to this code on your github fork so other will
> be able to make comments.
> I saw your latest code, but I would like to comment current code.
>

Host side:
https://github.com/balidani/JohnTheRipper/blob/bleeding-jumbo/src/opencl_rawkeccak256_fmt.c

Kernel:
https://github.com/balidani/JohnTheRipper/blob/bleeding-jumbo/src/opencl/keccak256_kernel.cl


I'll try to find the problem during the weekend.

Regards,
Daniel

[ CONTENT OF TYPE text/html SKIPPED ]

Powered by blists - more mailing lists

Your e-mail address:

Powered by Openwall GNU/*/Linux - Powered by OpenVZ