[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CABob6ipG4REgpjoGyK189ybycenVSxHs89yE54EQKf=2eKEKeg@mail.gmail.com>
Date: Sat, 8 Nov 2014 06:45:26 +0100
From: Lukas Odzioba <lukas.odzioba@...il.com>
To: john-dev@...ts.openwall.com
Subject: Re: Discuss Mask-mode with GPUs and changes required for
its support.
2014-11-05 16:02 GMT+01:00 Sayantan Datta <std2048@...il.com>:
> 1. Apply a default mask='?w' for formats supporting GPU generation when
> non-mask modes are used.
In my opinion to do it properly we should change all gpu formats to
expect passwords and salt in common format under some desired place.
The advantage of this I see is that we could get rid of some cpu-gpu
memcopy code for each format.
While we're on it we should design it in a way that will be efficient
for fast hashes, like overlapping it with kernel execution if
architecture allows for that.
Password preparation procedure should be still controlled by the
format, because of splitted kernels.
How to get results back is another question to ask here, since it will
influence changes in core - FMT_GPU_MASK is good idea.
I would leave results comparision on cpu for slow formats.
It would be good to have in mind a plan for multigpu usage before coding starts.
Lukas
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.
