Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Sat, 7 Jul 2012 20:39:03 -0400
From: Rob Fuller <jd.mubix@...il.com>
To: john-dev@...ts.openwall.com
Subject: Re: OSX problems with Jumbo-6 OpenCL

No joy:

# Raw MD5 Test

./john --format=raw-md5-opencl ../../rawmd5test.txt --device=1
OpenCL platform 0: Apple, 2 device(s).
Using device 1: GeForce GT 330M
Compilation log: Error getting function data from server
OpenCL error (CL_INVALID_PROGRAM_EXECUTABLE) in file (opencl_rawmd5_fmt.c)
at line (168) - (Error creating kernel. Double-check kernel name?)

 ./john --format=raw-md5-opencl ../../rawmd5test.txt --device=0
OpenCL platform 0: Apple, 2 device(s).
Using device 0: Intel(R) Core(TM) i7 CPU       M 620  @ 2.67GHz
Local work size (LWS) 1, Global work size (GWS) 2097152
Loaded 1 password hash (Raw MD5 [OpenCL])
Session aborted

# NT Test

 ./john --format=nt-opencl ../../nttest.txt
OpenCL platform 0: Apple, 2 device(s).
Using device 0: Intel(R) Core(TM) i7 CPU       M 620  @ 2.67GHz
Loaded 1 password hash (NT MD4 [OpenCL])
Session aborted

 ./john --format=nt-opencl ../../nttest.txt --device=1
OpenCL platform 0: Apple, 2 device(s).
Using device 1: GeForce GT 330M
Compilation log: Error getting function data from server
OpenCL error (CL_INVALID_PROGRAM_EXECUTABLE) in file (opencl_nt_fmt.c) at
line (176) - (Error creating kernel)


Is there a setting I can use to get better debugging data for you guys?

--
Rob Fuller | Mubix
Certified Checkbox Unchecker
Room362.com | Hak5.org


On Thu, Jul 5, 2012 at 1:53 PM, Solar Designer <solar@...nwall.com> wrote:

> On Thu, Jul 05, 2012 at 12:20:31PM -0400, Rob Fuller wrote:
> > I have symlinked /usr/local/cuda/lib to /usr/local/cuda/lib64 as well as
> to
> > /usr/local/cuda/lib/x86_64 the contents of which are:
> >
> > libcublas.dylib libcudart.dylib libcuinj.dylib libcusparse.dylib
> > libtlshook.dylib
> > libcuda.dylib libcufft.dylib libcurand.dylib libnpp.dylib
> >
> > Made the symlinks so I wouldn't have to keep editing the Makefile each
> rev,
> > but if there is a better way please let me know or if I don't have the
> > paths/links right, don't have something installed, etc..
>
> FWIW, on Linux with a default install of the SDK I do:
>
> export PATH=$PATH:/usr/local/cuda/bin
> export LD_LIBRARY_PATH=/usr/local/cuda/lib64
>
> and don't edit John's Makefile.  Maybe we should have this recipe
> somewhere in the documentation or/and on the wiki.
>
> Perhaps something similar would work for OS X as well?
>
> Alexander
>

Content of type "text/html" skipped

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.