I want to create/load pre-compiled opencl kernels for resnet50 model on Adreno GPU. I’ve built the tvm cpp rtvm deployment API and it is able to perform model inference. As per documentation, I can save kernel compilation time by saving opencl kernels in first run and then using the pre-compiled kernels later and this is done by passing "–pre-compiled parameter. But it’s not generating the opencl kernels binary. So how to generate the binary containing pre-compiled opencl kernels? Command used : ./rtvm --model=./ --device=opencl --pre-compiled=compiled.so --profile
