Parallel BFS schedule

juno_samples/rodinia/bfs/src/cpu.sch

@@ -48,8 +48,11 @@ let traverse_body = outline(inner);
 let (outer, inner) = fork-reshape[[1], [0]](collect);
 let collect_body = outline(inner);
 
+inline(bfs@cost_init, bfs@loop1, bfs@loop2);
+delete-uncalled(*);
 const-inline(*);
-simpl!(*);
 
 unforkify(init_body, traverse_body, collect_body);
-gcm(*);
\ No newline at end of file
+simpl!(*);
+gcm(*);
+xdot[true](bfs);