From the abstract:
We are investigating a new approach to compiling Standard ML (SML) based on four key technologies: intensional polymorphism , nearly tag-free garbage col lection [12, 46, 34], conventional functional language optimization, and loop optimization. To explore the practicality of our approach, we have constructed a compiler for SML called TIL, and are thus far encouraged by the results: On DEC ALPHA workstations, programs compiled by TIL are roughly three times faster, do one-fth the total heap allocation, and use one half the physical memory of programs compiled by SML of New Jersey (SML/NJ). However, our results are still preliminary we have not yet investigated how to improve compile time; TIL takes about eight times longer to compile programs than SML/NJ. Also, we have not yet implemented the full module system of SML, although we do provide support for structures and separate compilation. Finally, we expect the performance of programs compiled by TIL to improve signicantly as we tune the compiler and implement more optimizations.