Re: [osg-users] OpenCL: democracy for GPU computing?

2009-03-24 Thread Paul Martz
Interoperability (in terms of being able to copy data directly between
analogous OpenGL and OpenCL buffers) is a feature of the new OpenGL 3.1.

Paul Martz
Skew Matrix Software LLC
http://www.skew-matrix.com
+1 303 859 9466

-Original Message-
From: osg-users-boun...@lists.openscenegraph.org
[mailto:osg-users-boun...@lists.openscenegraph.org] On Behalf Of
hanne...@gmx.at
Sent: Thursday, March 19, 2009 7:21 PM
To: OpenSceneGraph Users
Subject: [osg-users] OpenCL: democracy for GPU computing?

http://www.behardware.com/articles/744-1/opencl-democracy-for-gpu-computing.
html

thoughts?
___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org

___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org


Re: [osg-users] OpenCL: democracy for GPU computing?

2009-03-20 Thread Paul Melis

Martin Beckett wrote:
Following OpenGL3 I'm not holding my breath for a standard developed by Khronos. 
  
It was actually developed by Apple, who submitted it to the Khronos 
Group (although they probably fiddled with the original spec, of course).

The bit about OpenCL not just being for GPUs but also for CPUs and Cell worries 
me - it sounds like a 'grand plan' - it supports manufacturers extentions so if 
you have an OpenCL app that needs Nvidia extentions why not just use Cuda?
Finally are there any real competitors to Nvidia in this space? Are you about 
to run heavy weight physics simualtions on an embedded Intel chipset?
  
Would you rather be forced to always use NVidia's stuff through Cuda 
than possibly have the option of using a different backend through OpenCL?
Only time will tell if there are going to be serious competitors, of 
course...


Paul
___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org


Re: [osg-users] OpenCL: democracy for GPU computing?

2009-03-20 Thread Robert Osfield
On Fri, Mar 20, 2009 at 4:57 AM, Martin Beckett m...@mgbeckett.com wrote:

 Following OpenGL3 I'm not holding my breath for a standard developed by
 Khronos.
 The bit about OpenCL not just being for GPUs but also for CPUs and Cell
 worries me - it sounds like a 'grand plan' - it supports manufacturers
 extentions so if you have an OpenCL app that needs Nvidia extentions why not
 just use Cuda?
 Finally are there any real competitors to Nvidia in this space? Are you
 about to run heavy weight physics simualtions on an embedded Intel chipset?


We'll have to see what happens with ATI's and Intel's OpenCL
implementation.  Supporting extensions/some code specifics is far better
than doing work having to do everything in completely proprietary API's as
more of your work will be reusable.

The OSG itself uses lots of OpenGL extensions but it hasn't become a great
problem for maintenance or development - the OSG still supports OpenGL 1.1
hardware as it does OpenGL 2.1 + extensions hardware.  This does suggest
that it won't be a show stopper that OpenCL has extensions and that we might
need to use them occassionally to get best out of particular hardware
platform.

Robert.
___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org


Re: [osg-users] OpenCL: democracy for GPU computing?

2009-03-20 Thread Martin Beckett
Yes I would much prefer a cross platform well supported API than be forced to 
buy NVidia (and have to tell my customers to).
But at the moment a lot of the OSG demos do not render properly on any other 
cards (at least on Windows) just because of the poor OpenGL drivers - so I'm 
not holding my breath for them to all correctly implement the OpenCL spec.

If GPGPU is going to become mainstream then we do need something like OpenCL 
(else DirectX will do it!). At the moment it doesn't matter - the application 
is generally so specialised that requiring a particular board isn't a problem.
But an industry wide spec for something which is technically complex and only 
used in niche areas needs a strong driving force behind it. OpenGL really took 
off because of SGI not a loose industry-wide steering group.

Still - you never know it might all work out :-)

--
Read this topic online here:
http://forum.openscenegraph.org/viewtopic.php?p=8862#8862





___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org


[osg-users] OpenCL: democracy for GPU computing?

2009-03-19 Thread hanne...@gmx.at

http://www.behardware.com/articles/744-1/opencl-democracy-for-gpu-computing.html

thoughts?
___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org