Where to find the OpenCL kernel implementation of the Operators?

In TVM DOC “https://docs.tvm.ai/tutorials/autotvm/tune_relay_mobile_gpu.html#sphx-glr-tutorials-autotvm-tune-relay-mobile-gpu-py”, it says that “The operator implementation for Mobile GPU in TVM is written in template form. The template has many tunable knobs (tile factor, vectorization, unrolling, etc). We will tune all convolution, depthwise convolution and dense operators in the neural network.”

My question is how can we add OpenCL kernel for custom operators aiming to running on Android? Could someone help me ?