clFlush() is taking more cycles.

Hi,

I am running Opencv with cl code on odroid-xu board, to measure performance of color conversion code, i have seen the difference when running normal code on cpu and opencl code on GPU, cpu is taking very less cycles compared to GPU, so i just measured cycles for each module in that, i found that clFlush() function is consuming more cycles. Is there any way to replace clFlush(), or to make it consuming less cycles.

Thanks,
Shabuddin.