[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <84fce73e910339b920766eb07fc78270@smtp.hushmail.com>
Date: Sun, 13 Jan 2013 03:14:20 +0100
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: Changes to common-opencl.c
On 13 Jan, 2013, at 2:48 , Lukas Odzioba <lukas.odzioba@...il.com> wrote:
> 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.
On the countrary, it ended up very well and most formats are using your shared function for finding out LWS! Actually I am not aware of any format that enumerates LWS itself - the ones not using your function are simply picking a fixed value that works fine in all situations. Maybe some of Sayantan's formats have a local function but that is because of his early multi-gpu support as far as I can tell.
Having said that, I have plans for using a private find_best_lws() for my split-kernel formats because they should do this enumeration with a lot lower iteration count than normal. I do GWS enumeration that way, and it speeds things up a great deal. Unfortunately I can't see any way to have a shared function support that.
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.
