Monthly Archives: September 2007

Mason found a problem with the Singleton Analysis Pass. Objects that were allocated and then written to an array were not correctly flagged as non-singleton. Besides this rather obvious bug I also fixed the way context slots are handled. When … Continue reading →

I finished today the Singleton Allocation Analysis phase which was the last building block missing to enable Invariant Allocation Hoising (move NEWs out of loops, which is our equivalent of escape-analysis based stack allocation). The SAA pass goes over the … Continue reading →