[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 03 Jul 2013 11:57:48 +0000
From: "johnfoobar" <johnfoobar@...mail.org>
To: john-users@...ts.openwall.com
Subject: Re: mscash2-opencl failing device tuning (was: use of
undeclared identifier 'k')
> On 2 Jul, 2013, at 1:22 , magnum <john.magnum@...hmail.com> wrote:
>> On 30 Jun, 2013, at 19:23 , johnfoobar <johnfoobar@...mail.org> wrote:
>>> ./john --format:mscash2-opencl -t
>>> OpenCL platform 0: NVIDIA CUDA, 1 device(s).
>>> Device 0: Quadro 2000M
>>> OpenCL error (CL_OUT_OF_RESOURCES) in file (common_opencl_pbkdf2.c) at
>>> line (433) - (SYNC FAILED)
>>
>>
>> It looks to me the format's auto-tuning to device fails. Try using -t=0
>> instead of just -t for a really small workload (and thus poor speed).
>> Does it work then? If it does, you can experiment with different values
>> for local and global worksizes like this:
>>
>> LWS=64 GWS=512 ../run/john -t -form:mscash2-opencl
>>
>> If that too works, try doubling the GWS figure until you find the
>> fastest value that works fine. Please report back.
>
> I was just made aware you are using the unstable branch, which oddly
> enough is severly outdated. Please use the bleeding-jumbo branch instead.
> Chances are the problem is already fixed.
running
john -t -form:mscash2-opencl
using
bleeding-jumbo (1630b9b9c2f227fdb67dc570b7693f00a70092a5)
works without any issues for me.
In case anyone wants to know how fast that gpu can crack that hashtype:
Device 0: Quadro 2000M
Optimal Work Group Size:512
Kernel Execution Speed (Higher is better):0.102763
Optimal Global Work Size:24576
Benchmarking: mscash2-opencl, M$ Cache Hash 2 (DCC2) [PBKDF2-SHA1
OpenCL]... /
DONE
Raw: 6660 c/s real, 6678 c/s virtual
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.
