Polymorphism is a key feature of modern, strongly-typed programming languages including Ada, C++, and ML. This research program investigates a new approach to implementing polymorphism, based on associating type information with each activation of a polymorphic function, rather than with data. This approach permits type information to be utilized effectively at runtime or link-time, not just at compile-time. It enables compilers to generate polymorphic object code from parametrically polymorphic source code, while still using specialized, space-efficient data formats that can be shared with other software components. The approach also supports a large variety of powerful non-parametric features in the source language and in the development and runtime environments. The focus of the research will be on generating prototype compiler implementations and language extensions, that can be evaluated using performance benchmarks. The underlying aim of the research is to make polymorphic programs easier to develop and to integrate into existing systems, thereby, hastening the acceptance of modern languages by industrial and commercial users as well as by academics.
