Chapter 18 Excerpted from the 1994 Research Summary 
(Copyright 1994, UC Regents)

Contact person: Carol Block, ILP Coordinator 
(ilp@hera.eecs.berkeley.edu)
510.643-6691, 643-6694 fax


============
INTRODUCTION
============

The development of high-quality software is a difficult, multifaceted
problem.  Our research in software systems aims to advance
state-of-the-art software development. The areas we are studying
include languages used to specify computations, interactive
environments for the development, description, and maintenance of
software, methods for translation of higher-level languages to target
machine code, runtime systems to provide efficient execution and
debugging, and techniques to teach and support problem-solving in
programming.  Beyond the traditional attention to sequential
computing, our research increasingly focuses on massively parallel
computing.

We are designing and building systems to support the creation of
software and the communication of ancillary technical
information. Ensemble, an outgrowth of the language-based Pan editing
system and the VorTeX document-development system, integrate
facilities for construction and presentation of text, programming
languages, and multimedia documents.  Ensemble uses programming
language notions to formally specify document structure and
appearance; high-quality formatting enhances the understanding of
program presentations. To speed the development cycle for programs,
formatted text and other document types, Ensemble uses an incremental
approach. Syntactic and semantic analysis, along with translation of
the document, are dynamically recomputed only for the edited portions.

Our research in Ensemble includes work on general-purpose modules such
as the Cleaver system for specifying incremental compilers, and the
Colander system for static semantics checking, as well as specific
incremental analyzers for C and Fortran.

A major theme of our programming language research is the expression
of parallel computation at a high level, thereby freeing the
programmer from details of the machine architecture. Ongoing work in
this area includes FIDIL (a language for scientific applications),
pSather (a parallel object-oriented language), ID (a non-strict
implicitly-parallel language), and a version of Lisp with extensions
for parallel processing.  Since parallel processing applications
require performance, high-level languages must provide expressiveness
without sacrificing efficiency. At a lower level closer to the
machine, Split-C provides a small set of extensions to C and serves as
both a user language and a compiler target.

Our parallel language efforts also encompass compilation and
optimization of programs, as well as the design and implementation of
runtime systems. For parallel compilation, we study the advanced uses
of symbolic analysis, type analysis, dependence analysis, and
annotations to translate and optimize explicitly-parallel
code. Runtime systems under study include basic services such as
lightweight thread scheduling, load balancing, global naming, and data
structure libraries such as Multipol, the pSather libraries, and the
Lisp toolbox. Our research necessarily involves a significant effort
in parallel applications development, often in collaboration with
researchers from other departments.

In the area of teaching, software development requires programming and
problem-solving skills attainable only by a significant educational
investment. Our research in developing effective methods for teaching
programming and reasoning skills incorporates problem-solving methods
used by experts with an interactive Hypermedia form for supporting
instruction.