The goal of this project is to develop techniques and language abstractions for efficient parallel computing on multicore computers.  Current work includes the design, analysis, and implementation of techniques for parallel task creation and scheduling, as well as general-purpose abstractions for expressing massively parallel programs.  Our approach is to understand the underlying architecture well by identifying the magnitude of "constant factors"  and to take these factors into account in our design and implementation.  We believe that this is important to ensure efficiency in parallel computing.   We are currently developing a C++ library based on this work.