[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <4F60F934.8090905@gmail.com>
Date: Wed, 14 Mar 2012 17:01:56 -0300
From: Claudio André <claudioandre.br@...il.com>
To: john-dev@...ts.openwall.com
Subject: Re: New function in common-opencl
I had problems with default KPC on CPU. I suppose this is annoying you.
Using the information provided by CL_DEVICE_MAX_COMPUTE_UNITS and
CL_DEVICE_TYPE is possible to improve this.
I'll try something.
Claudio
Em 14-03-2012 15:54, magnum escreveu:
> On 03/14/2012 07:27 PM, Claudio André wrote:
>> Hi, i need a new function to check if LWS is ok in the running hardware
>> (need it to optimize).
>>
>> I put in common-opencl (seems the best place to be). Any complains?
>>
>> Complete patch attached if someone wants to see it.
> I have no idea if that exact code is good or bad, but I like the general
> idea of adopting to actual hardware. We need to make different decisions
> (default KPC in particular) for CPU vs GPU, and also based on number of
> cores reported.
>
> BTW, does anyone know why the (few, all nvidia) GPU cards I've seen seem
> to report one eighth of the actual number? For example, a "240 core"
> GTX280 reports "Parallel compute cores: 30".
>
> magnum
>
Content of type "text/html" skipped
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.
