OpenGL® 3.1 w/ GLSL? 1.40, OpenSL ES? 1.0 and much more
http://www.khronos.org/
It's a great news to see OpenGL® 3.1 w/ GLSL? 1.40 come out so soon after OpenGL® 3.0 was shipped and panned, not to forget that OpenSL ES? 1.0 came out as well.
It's a great news to see OpenGL® 3.1 w/ GLSL? 1.40 come out so soon after OpenGL® 3.0 was shipped and panned, not to forget that OpenSL ES? 1.0 came out as well.
Comments
I wonder when we can reasonably expect to actually see this available in OS X though? It would be nice if we had access to the 3.1 drivers on day one like Windows and Linux.
I know the older legacy cruft has been deprecated. There seems to be palpable amount
of relief on the OpenGL forums about 3.1 coming so soon.
Good job Khronos!
It's nice to see Khrono's getting 3.1 done so quickly. I think that goes a long ways towards removing the feeling that OpenGL was dying.
I wonder when we can reasonably expect to actually see this available in OS X though? It would be nice if we had access to the 3.1 drivers on day one like Windows and Linux.
Since OpenGL 3.1 works with OpenCL 1.0 you'll see both in 10.6 out-of-the-box.
Wait, are you insinuating that OpenGL CL 1.0 needs OpenGL 3.1?
What's OpenGL CL 1.0?
At any rate, OpenGL 3.1 leverages OpenCL 1.0, not the other way around.
Check the link for the vids on how OpenCL improved performance.
What's OpenGL CL 1.0?
...at typo. I wasn't sure what you meant and was going to write "Wait, are you insinuating that OpenGL needs OpenCL?" but then thought you meant the reverse and didn't change OpenGL to OpenCL before posting.
At any rate, OpenGL 3.1 leverages OpenCL 1.0, not the other way around.
Can you send me a link that clearly shows OpenGL 3.1 leveraging OpenCL?
Perhaps I should also be paging Brad Oliver. Brad? Are you still around?
http://www.cgdigg.com/story.php?titl...physics-at-gdc
Check the link for the vids on how OpenCL improved performance.
There was a video shown at Engadget too with cloth dynamics:
http://www.engadget.com/2009/03/27/h...retty-dresses/
What someone had commented on was that the demo was done with AMD not Intel even though Havok is a subsidiary of Intel now. Kinda strange.
Anyway, I was curious to see the performance of the 9400M GPU for this type of stuff so I ran some of the CUDA demos from Nvidia. They are pretty amazing. The 9400M in the Mini has 16 x 0.3GHz cores, which is pretty low but you can still see applications where it greatly outperforms the CPU. It can apparently be overclocked (or I should probably say de-underclocked):
http://www.hardmac.com/news/2009/01/...ook-on-windows
EDIT: don't use this tool. The 9400M in the Mini and MBA is underclocked to 350MHz from 450MHz core and the shader clock is 800MHz instead of 1.1GHz but running the test for the best value in this tool caused a BSOD and subsequently some graphics corruption occurred. Uninstalling it and rebooting fixed it. Apple has actually put the 9400M at 450MHz in the Macbook, which is odd because it's a laptop so would benefit more from being underclocked.
The CUDA kit included a particle smoke demo that is like a rocket with smoke trails and they have a light casting soft shadows onto the ground and even on the 9400M, it runs at 10-15fps. This seems slow but when you consider trying to do this on the CPU using 3D software, you'd be hard pushed to manage 1 frame rendered in a second for the most basic stuff so even a low end GPU can do certain things up to 10 times faster. It's all about how the code is optimized and the improvements will be great when developers get things done right. Using the CPU and GPU together is where you'll see the biggest steps forward.
There are limitations though because when you run the high end CUDA demos, the interface stutters all over the place so dragging windows and things are slower when using the GPU for computation. To make matters worse, when I ran the Mandelbrot app, it actually crashed at one point and the entire OS graphics locked up. Even hitting sleep put the machine to sleep but the image stayed on screen. A reboot was needed.
I hope that OpenCL gives some level of control over crashing the way that OS X does for the CPU so that apps that do crash can be closed without a reboot and also I'd hope that there can be reserved allocations for the GPU usage so that the main interface doesn't slow down to the point of being unusable. I wouldn't say that it became unusable running the CUDA demos but it slowed down noticeably.
This is where machines like the Macbook Pro will be better as the CUDA code can run on the dedicated card while the interface runs on the 9400M as long as support is added to do this. This would mean that even if the 9600M GT app crashed, it wouldn't take out the OS GUI.
Nvidia considers porting PhysX to OpenCL
Nvidia is clearly still committed to accelerating PhysX on GPUs via CUDA in the near future, but it looks as though a port to OpenCL is a very real possibility. Should Nvidia port PhysX over to OpenCL, or would that result in unfair quality-control complications? Let us know your thoughts in the forums.
http://www.bit-tech.net/news/hardwar...sx-to-opencl/1