Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Wed, 11 Dec 2013 02:07:58 +0100
From: magnum <>
Subject: Re: md5crypt-opencl

On 2013-12-11 01:23, Lukas Odzioba wrote:
> 2013/12/11 magnum <>:
>> Thanks, this is committed in 021717e. I also enabled all test vectors in
>> ce36518 - in CUDA format too (its kernel probably needs a similar overhaul).
> Thanks, just for the reference I am posting more test results:
> all test vectors:
> 570@...l: 1139K c/s
> 7970@...l: 1516K c/s
> {"$1$salt1234$u4RMKGXG2b/Ud2rFmhqi70", "abcdefgh"}:
> 570@...l: 1255K c/s
> Tahiti@...l: 2293K c/s
> I really like CUDA api, but you know... let's say that making it
> faster on amd has higher priority than making it equally fast using
> different command.

Maybe I can take care of the CUDA format (mostly just mimicing your 
OpenCL optimizations) - I've had plans to "concentrate" just a little on 
CUDA in near future to get the core/shared stuff to same state as OpenCL 
(mainly multi-device options parsing and fork support).


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.