Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Thu, 24 May 2012 17:55:26 +0400
From: Solar Designer <>
Subject: Re: Cuda version of John the Ripper

On Thu, May 24, 2012 at 08:50:06AM -0400, Cody Sapp wrote:
> Okay that worked, but when I ran make linux-x86-64-cuda (since my computer
> is 64-bit), these last few lines appeared before the command stopped
> executing:
> t -ldl -lcudart -o ../run/john
> /usr/bin/ld: cannot find -lcudart

It's a bit puzzling that you got this far since the error indicates that
your CUDA SDK is not fully available for this build.  Have you used CUDA
before?  On this system?

Here are settings that work with a default install from

export PATH=$PATH:/usr/local/cuda/bin
export LD_LIBRARY_PATH=/usr/local/cuda/lib64

What's your reason to use a CUDA-enabled build of John the Ripper?  What
hash types do you intend to use it with?  Do we even have CUDA code for
those?  (It's not like a "CUDA version of John the Ripper".  It is CUDA
code for a handful of individual hash types.)

Have you used John the Ripper before?  Do you currently have it working
on your target hashes?  What performance are you getting, which you're
presumably hoping to improve by using CUDA?  Are you already using
OpenMP or another approach to parallelization?  What CPUs and GPUs do
you have in that machine?

Is this all just for the sake of it - try out a CUDA-enabled password
cracker against whatever hashes and in whatever way?  If so, you're
probably equally interested in OpenCL as well (it is similarly supported
for individual hash types), are you?

> Are those errors going to be problematic?

Of course.  Your build of JtR has not completed.


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.