[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sun, 13 Jan 2013 12:23:39 -0200
From: Claudio André <claudioandre.br@...il.com>
To: john-dev@...ts.openwall.com
Subject: Re: Changes to common-opencl.c
Em 12-01-2013 23:48, Lukas Odzioba escreveu:
> In my opinion we should have some kind of "framework" that can handle
> multigpu support itself. Some time ago I tried to move function
> finding "best" GWS/LWS to common-opencl but it didn't ended well and
> now we still have bad situation where almost each format have a lot of
> copy pasted code for doing that. When it comes to multigpu there will
> be more copy pasting and this is not a good thing. Ideally we should
> be able to run code with different in speed gpus and have almost no
> speed regression (I mean that sum of speeds from testing gpus
> separately should be equal multigpu mode). This is not easy at all
> especially in one thread, but less problematic than the same thing in
> cuda. Maybe it will require some changes in fmt_ interface, or could
> be done by some "task manager" working inside crypt() function to
> divide data and gather results from gpus.
I agree, but I need to have a good undestanding about what is really
necessary to suggest / change. Now, only Sayantan produced muti-gpu
code. So, I intend to create the possibility and latter create something
multi-gpu in order to see what is possible/necessary.
My fellings is that without core modifications we will have some speed
regression (as you said in [1]). Better than nothing but not perfect.
For now, I'm only handling better a multi-gpu environment. Using only
one thread, the format creator still has to figure out which LWS/GWS
pick in order to result in [1]. Maybe using Sayantan work and your idea
in opencl_find_best_workgroup() is enought. Now, I have no idea if it is
true.
[1] - "have almost no speed regression ('you' mean that sum of speeds
from testing gpus separately should be equal multigpu mode)"
>
>> Problems:
>> - hot-plug is not going to be handled. Well, it is not handled now too.
> What do you mean by hotplug? Adding gpus while john is runnig?
Exactly.
Claudio
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.
