Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Tue, 21 Feb 2012 22:39:25 +0100
From: Lukas Odzioba <lukas.odzioba@...il.com>
To: john-dev@...ts.openwall.com
Subject: Re: OpenCL vs CUDA

2012/2/21 magnum <john.magnum@...hmail.com>:
> Is this a general difference you see too?
It is "normal". OpenCL code for md5crypt is slower for opencl, it is
because of lack of shared mem in opencl.

For phpass on nvidia, cuda and opencl should have comparable speed (
not using shared only registers).
For  phpass on amd opencl is over 3x faster than nvidia card for the
same price (gtx460-300kc/s 5850 ~950c/s).
In my opinion we should have cuda build optimized for nvidia cards,
and opencl build optimized for amd's, but still working if possible on
nvidia's. This might require common set of available formats for cuda
and opencl build.

Lukas

Powered by blists - more mailing lists

Your e-mail address:

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.