The support of out-of-order mode ?


The out-of-order execution is one of the key features of OpenCL. However, I can only get it work using Intel’s implementation. And there is no difference from in-order execution using Nvidia’s implementation. See my post on Nvidia’s OpenCL forum.
Can anyone state the support status of out-of-order execution on popular platforms?


Is there a particular reason why you think you would benefit from out-of-order execution?

Currently as a beginner I am learning OpenCL. So I want to try the features listed in specification. And you are from Khronos working group, you should know out-of-order mode is a powerful tool to execute commands more efficiently by filling the time gap. At least it is necessary to clarify the support status of OpenCL.

It’s optional and some (most) vendors don’t support it.

More widely supported is using multiple in-order command queues (and events for synchronization).

Nvidia Fermi Gpus has out-of-order execution on thread level … epaper.pdf

If i can simulate 2000 molecules with my pentium-centrino 2GHz, what can do with an AMD HD 7850 ?

Calculation time complexity is nearly linear with molecule numbers.