How to extend NNVM/TVM as a static compiler for HW accelerator like NVDLA?

Questions
10

Maybe we can have a try on chisel-nvdla first.