OpenCL API overhead

Hi ,

There are some profile API could let programmers understand the time (kernel execution , submit task to queue ... )

Does Altera provide the GUI tool (such as AMD CODEXL , NVIDIA nvvp ) ?

Because I write a program which enqueued > 1000 times.

But I estimated the time about write buffer , read buffer , enqueue (queued , summit , start , end) , I still lost some time .

For example :

kernel time : 30 s (not include the CPU time) - I used time API(C library)

The time I got through OpenCL API is 20 s

=> 10 s is lost ...

Thanks