opencv3.2 umat is slower than mat...

edit

Hi, i want to compare the performance of using OpenCL with the case of not using it. When i use DGpu device(NVIDIA GTX1050), it records fastest time to execute gaussian blur and canny edge detection. But, the problem is when i use IGPU(Intel HD Graphics 630) and CPU device with UMat, it is slower than the case using Mat type. is it normal that using opencl is slower than the case not using it?