What to do with CLK_DEVICE_QUEUE_FULL?

When I use enqueue_kernel on gpu device then CLK_DEVICE_QUEUE_FULL happens on a small amount of items(240+ on intel graphics) which makes dynamic parallelism of opencl 2.0 is quite useless. Is there any good way to deal with it?

Program hangs when I use approach - while(0 != enqueue_kernel(…


This sounds like a similar question that was recently asked on the Intel OpenCL forum:


If so, let’s continue discussion there, thanks!