gam wrote:Maybe this new general-purpose parallel programming language would be useful to accelerate NGD
on modern multi-core and stream gpu processors? Looks like the holy grail to me.
Check it out here:
http://www.khronos.org/opencl/Greg
I would not hold my breath waiting for the Khronos group to come with practical solution to anything.
The Khronos group team is like dishonest politicians of the software industry, you the king that always jumps in front of a parade to take control of what other people are doing and take credit for it.
Openly was very good up until the specifications were made by the OpenGL Architecture Review Board, but ever since the big companies like Microsoft and Silicon Graphics left the committee because of the anarchy create by a whole lot of other companies the API because stagnant but stable.
But when the Khonos group took over just ruined the specifications.
This is also the group the came up with Collada, perhaps the worst, more pedantic and useless format I have ever seen.
The only thing you can expect from the incompetents at Khronos group is they taking good project by other people ruined and then calling it industry standard.
Knowing this is coming from the Khronos group I and not going to even bother reading it, I am sure it will promise wonders in the draft but it will deliver very little.
What hardware had the Khronos group made so that they can take the task of creation a multicores language? What do they know about the internal of any OS?
What do they know about the Cell, the Xenon, the Intel I7, the new GPU, etc.
What they know is the same thing you and I know if you are working with any of this platforms.
I am no particularly happy with the current way threading is handled with the current OS like XP, vista, Mac and Linux.
The problem with multi cores is that Microsoft created the thread API for windows and the open source community made pThread in response, as usual pThread is much more inferior to native windows treading but the is what you get wit t he group take a take they are no qualified to take.
A new threading API with much lower latencies is need it, the OpenMp turn out to be inadequate for many type of tasks, plus it is not batter than native threading because it is buid on top of the current threading interface.
What I can tell you is that the Khronos group in no way shape or form has the qualification to take on a task like that.
A task like that is better done by a real company like Microsoft or Intel, no by the moronic groups that are only in the way of everybody else.
Let the Khronos group fu*k themselves and take somebody else for a fool, they are not take any of my time anymore.
I can guarantee you that these OpenCL is the open source version of Cuda, only way more medicre and ineficient. They cannot see a project they cannot plagiarize.