Hi everyone,
I want to know where TVM specifies the dependencies of kernels that run on its target. For example, assume that we have a ML model and build it by TVM for an OpenCL device target. Now, my question is how TVM handles the dependencies of these kernels in code.
I checked the codes of OpenCLModuleNode::GetFunction and its usages in higher levels but I could not find my answer. It seems that it is handled in somewhere else. Is there anybody knows where it is handled?
Best,