Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 02 Aug 2012 20:23:41 +0200
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: wpapsk error report

On 2012-08-02 19:57, Frank Dittrich wrote:
> On 08/02/2012 06:40 PM, Lukas Odzioba wrote:
>> Some time ago I got the following error report for wpapsk-cuda format.
>> Has anyone have access to similar gpu (GeForce 8600GT) and can try
>> reproduce this bug for the latest and this: b6b881d magnum jumbo
>> version?
> 
> I don't, but:
> Would it help to include CUDA / OpenCL specific version info into
> --list=build-info?
> 
> Is there also a run-time version which can differ from built info?

There's even the issue that you can have eg. OpenCL 1.2 compatible
*drivers* at run-time but some or all of your *devices* are OpenCL 1.1
or 1.0. This can currently be seen with --list=opencl-devices. For
example, any card not physically capable of byte-addressed stores will
never get better than OpenCL 1.0 regardless of driver updates (well
unless they work around it in software, but there's no chance they would
do that for an old card).


> And is there a way to include driver version in run-time-info?

You could re-use some of the --list=(cuda|opencl)-devices code and put
this stuff into --list=build-info if you want to, but I would personally
not bother. I do not object either, but we should try to use shared
functions eg. in opencl-common.c so future fixes don't have to be
applied to several places.

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.