Removing Implicit Deadlocks on a Thread-per-core Architecture with 2-phase Processing