Several tests show that the number of cycles of a test circuit has increased, sometimes dramatically.
What are the differences?
swap and cnot are not decomposed: this makes resource use estimation overly conservative and thus the circuit more sequential; I estimate that this make cost around 10% more cycles, so it cannot explain everything
the gates entering the mapper are ordered by criticality; that is supposed to be better most of the times
Several tests show that the number of cycles of a test circuit has increased, sometimes dramatically. What are the differences?
This must be investigated!