[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <d8d2adb894c52580c2a4c77f3b6fd4eb@smtp.hushmail.com>
Date: Sun, 12 Feb 2017 18:40:31 +0100
From: magnum <john.magnum@...hmail.com>
To: john-users@...ts.openwall.com
Subject: Re: OpenCL CL_INVALID_DEVICE error in
common-opencl.c:2061 - Error querying CL_DEVICE_TYPE
On 2017-02-10 15:38, Solar Designer wrote:
> JtR doesn't automatically use multiple GPUs. You'll want to use
> command-line options like "--dev=0,1 --fork=2".
>
> More importantly, for fast hashes like LM it is pointless to use a GPU
> when you're not using "--mask". So you should either use "--mask" on
> its own, or along with another cracking mode (referring to that other
> mode's parts of the candidate passwords with "?w" in the mask).
>
> magnum -
>
> Maybe JtR should print a warning (and what to do) whenever a fast hash
> OpenCL format is invoked and "--mask" is not specified? Otherwise every
> new person trying to use JtR on GPU against a fast hash would first fall
> into this trap.
I'm not sure... I dislike warnings thrown at me when I do know what I am
doing.
> Similarly, I think JtR should at least warn, or better yet refuse to
> run, when the list of devices given is longer than the number of forks.
Perhaps using eg. -dev=all or -dev=0,1 should imply fork as needed?
There might be some caveats with MPI though.
magnum
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.
