Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 05 Jul 2012 11:50:42 -0300
From: Claudio André <>
Subject: Re: common find_best_workgroup()

Em 05-07-2012 11:23, magnum escreveu:
> I have already committed an experimental fix. Seems to work fine but 
> still testing (I also changed all formats except Claudio's so they use 
> the shared function) 
I can change mine too. I need only two "new" things inside the default 
find_best stuff.

I will comment on one of your commits, just have to  check if it makes 
sense on AMD first.

> I'm starting to think we should drop the whole idea of
> find_best_workgroup. I mostly see random results (picking totally
> different values at different runs). For most kernels you will be able
> to pick a good default. The more critical one is global_work_size, and I
> believe that too can have a good default per format.
> magnum

We have been getting different results using "find_best" functions. It 
is something i really enjoy to use (on new hardware AND on 
experimentation). However, i agree that it is becoming clear it is not 
really necessary. I mean, the max possible value is always a good choice.

Anyway, i would like to see if a bug inside AMD driver for 7970 is fixed 
and than drop find_best_workgroup .


PS: i never use "find_best_workgroup" on fast hashes.

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.