CFA -- Workshop on Compilation of Symbolic Languages

                            --- Call for Workshop Participation ---


                                Compilation of Symbolic Languages
                                            for Parallel Computers
                                            October 31--September 1


                                            in conjunction with
                          International Logic Programming Symposium




                              Ian Foster (itf@mcs.anl.gov)
                              Math & Computer Science Division
                              Argonne National Laboratory
                              Argonne IL 60439, USA


                              and


                              Evan Tick (tick@cs.uoregon.edu)
                              Dept. of Computer Science
                              University of Oregon
                              Eugene OR 97403, USA




We invite all researchers actively studying compilation techniques
for parallel, symbolic languages, to participate in this workshop.


The purpose of this workshop is to bring together researchers working
on the compilation of logic, functional, and procedural programming
language for parallel computers. Each of these communities is faced
with the problem of inferring efficient memory management and
scheduling decisions. Each has made some progress, primarily on
shared-memory parallel computers. In the process, many techniques
have been discovered, developed, and rediscovered. This workshop is
meant primarily as a forum for the discussion of these techniques and
surveys of relationships between similar techniques developed in
different communities. A secondary role is to stimulate
discussion of the developments required to permit effective
compilation for large-scale, distributed memory parallel computers.


The fundamental problems facing researchers interested in compilation
of logic, functional, and procedural programming languages for
parallel computers are essentially the same. However, differences in
the basic programming paradigms have led to the communities
emphasizing different aspects of the parallel compilation problem.
For example, parallel logic and functional languages provide
dataflow-like formalisms in which control dependencies are
unimportant. Hence, an important focus has been on techniques that
try to infer when sequential control flow can safely be imposed.
Granularity analysis for scheduling is a related problem. The
single-assignment property (central to the dataflow model) leads to a
need for analysis of memory use in order to detect opportunities for
reuse. Much of the work in each of these areas relies on the use of
abstract interpretation techniques and similar methods.


In contrast, research in procedural languages has emphasized the
problem of inferring data dependencies in order to determine when
sequential control flow can safely be relaxed. A related area of
research is the automatic partitioning and distribution of data
structures. This topic has not been addressed in the logic and
functional programming communities but is important on large-scale
parallel computers.


On large-scale parallel computers, the difficulty of automatic
inference of good schedules has led to an interest in application-specific
compilation. In this approach, a program is automatically restructured to
produce a new program that will apply some programmer-specified scheduling
strategy on a particular parallel computer.


There is clearly both a commonality of interests between researchers
in these different fields, and large differences as to emphasis and
techniques. We hope that this workshop will encourage at least an
appraisal of the fundamental problems. We should point out that an
understanding of compilation techniques developed by different groups
is also important when programming paradigms are combined. For
example, compilers for logic languages that permit incorporation of
functional or procedural components can benefit from information
provided by functional and procedural compilers.


We solicit papers on compile-time implementation issues related directly
to parallelism: i.e., static analysis and restructuring techniques
allowing more efficient parallel execution. This workshop does not
cover parallel architectures or the implementation of parallel systems.
Submissions in these areas will be directed to the corresponding
workshops, and may be presented in overlapping sessions.


The fundamental problems in parallel compilation are independent of
particular languages and programming paradigms; we hope that this
workshop will encourage an appraisal of these problems and the
recognition of common solutions.


Please send a 1--3 page abstract of research to tick@cs.uoregon.edu
or itf@mcs.anl.gov by September 13, 1991. Include your address,
e-mail address and work telephone number. Notification will be given
soon after.
--

Return to the comp.compilers page.
Search the comp.compilers archives again.