I am pleased to report that I have certified the 5th increment of Co-dfns today. This increment brings the ability to actually run some real benchmarks on the compiler, specifically the Black Scholes benchmark mentioned in my ARRAY '14 paper.
This certification took longer than expected for a variety of reasons, but I have obtained some very useful data from the process which I think gives me a clear picture of how to improve the process and reliability of the code moving forward. This improvement should also net some definite productivity gains, since most of the loss in productivity when developing Co-dfns occurs during certification bog downs, rather than the development of the code itself.
Co-dfns is still a private piece of software at the moment, but interested parties should contact me.