Build Halide code for OpenCL GPU

[ArunabhaDeb2018]

Hello all,
I was trying to build halide code for opencl gpu device. I was able to build and run the code for the gpu target after going through the halide tutorials for gpu.But while debug is enabled to check which kernel is consuming how much latency,I could see compilation of opencl kernels (which happens everytime i run the code) is taking considerable amount of latency. My halide code is little complex and it takes more than 2s each time to compile the kernel(according to the debug log ).
The problem is every time I call my halide function from CPU wrapper side, halide is taking more than 2s to compile the halide kernel.
Is there any way to do it only once and save the kernel so that for next runs halide does not compile the kernel again?
Thanks in advance.