Refactor the compile engine into a cleaner interface.

Duplicate the CompileEngine interface.

Refactor the graph_runtime_codegen to invoke the new LowerTE pass

More changes

Things appear to be working

Some tracing to get Relay code to flow through too.

Disable some assertions as exp.

Tweak printing for now

Fix a few bugs: (#13)

1. Don't add relay main function to list of lowered TIR functions
2. Don't skip visiting call to relay function in graph runtime codegen

Remove debug prints.

Start refactoring

Split out shared data structures

Fix implicit duplicate decl of IsDynamic

Clean up handling of name + global prim fn

Clean up the code and debug issue introduced by previous hack

Clean up the debugging

Do C++ lint clean up

src/driver/driver_api.cc

@@ -246,14 +246,17 @@ std::pair<IRModule, IRModule> SplitDevHostFuncs(IRModule mod_mixed, const Target
   }
 
   if (target->kind->device_type == kDLCPU && target_host == target) {
-    ICHECK(mdevice->functions.empty()) << "No device code should be generated when target "
-                                       << "and host_target are both llvm target."
-                                       << "\n";
+    // We need to relax this check for just TIR functions.
+    // ICHECK(mdevice->functions.empty()) << "No device code should be generated when target "
+    //                                   << "and host_target are both llvm target."
+    //                                   << "\n";
   }
 
   return {mhost, mdevice};
 }
 
+// Can we make this take one annotated IRModule?
+//
 // Build for heterogeneous execution.
 runtime::Module build(const Map<Target, IRModule>& inputs_arg, const Target& target_host_arg) {
   auto pass_ctx = transform::PassContext::Current();