Programming Parallel Algorithms

                           Guy E. Blelloch
                      Carnegie Mellon University

In this talk I will discus how data-parallel languages along with the
work-time paradigm (for deriving complexities) are very well suited
for describing PRAM-like algorithms.  In particular I will describe an
experimental language, NESL, and show several examples of algorithms
written in the language.  The language has several features that make
it particularly well suited for implementing parallel algorithms,
including the ability to use nested structure and nested parallelism
(not offered by other data-parallel languages) and a formal way of
deriving complexity (work and steps) and then mapping it onto PRAM
complexities (processors and time).