An "HPspmd" Programming Model

Bryan Carpenter, Geoffrey Fox and Guansong Zhang

The poster motivates and discusses some features of a language
model combining characteristic data-parallel features from HPF
with an explicitly SPMD programming style.  This model, which
we call the "HPspmd" model, is designed to facilitate direct calls to
established libraries for parallel programming with distributed data.

The slow uptake of HPF can be partly attributed to
immaturity in the current generation of compilers, but it seems likely
that many programmers are actually more comfortable with the Single
Program Multiple Data (SPMD) programming style, perhaps because the
effect of executing an SPMD program is more controllable.
Many higher-level parallel
programming environments and libraries assume the SPMD style as their
basic model, including ScaLAPACK, PetSc, DAGH, Kelp, the Global Array
Toolkit.  These successes aside, the library-based SPMD approach to
data-parallel programming certainly lacks the uniformity and elegance
of HPF.

The class of programming languages we discuss borrow certain ideas
and compilation techniques from HPF, but relinquish some of its
principles.  We retain is an explicitly MIMD (SPMD) programming
model complemented by syntax for representing distributed arrays, with
syntax for expressing that certain computations are localized to certain
processors.  The claim is that these features will make calls to various
data-parallel libraries about as convenient as making a call to an array
transformational intrinsic function in Fortran 90.

Early experience with a Java-based HPspmd language called HPJava
will be related.