nvdla/sw

what's the diffenence between dependency order and topological order?

Opened this issue · 0 comments

why the compiler pre pass( splitNodes/fuseSubEngineOps...) use graphs.dpendecyOrder, while the post pass(generateDependecyParams/resovleMemory) use topological_order generated by topological sort?
what's the diffenence between dependency order and topological order?