You need to change the recorded kernels to others

There is a computational task written in Golang, which is now considered on CPU cores, it needs to be redone so that it, in the same logic of breaking down into subtasks, is considered on the GPU. I don’t understand how to do it, please help

Look for a Go package to interface with CUDA. The Go compiler can target WASM, but that is unlikely to give you the computational power that you seek by moving to a GPU.