The SPC Project

Description

In automatic, retargetable compilation low-cost, analytic cost estimation techniques are crucial in order to efficiently steer the optimization process. Programming models aimed at optimum expressiveness of parallelism, however, are not amenable to static cost estimation. We present a new coordination model, called SPC, that imposes specific restrictions in the synchronization structures that can be programmed. Imposing these restrictions enables the efficient computation of reliable cost estimations paving the way for automatic optimization.

There a number of papers [1,4] that illustrate the advantages of SPC's symbolic performance predictability with respect to optimizing compilation of which the process-algebraic framework is currently under development (Frits Kuijlman). In addition, we are looking at the advantages SPC's synchronization restrictions offer with respect to scheduling in general (Andrei Radulescu). This subproject is in collaboration with Henk Sips (see the Automap project).

Regarding SPC's limited expressiveness we are investigating the conjecture [1] that the loss of parallelism when programming in SPC is typically limited to a constant factor of 2 compared to the unrestricted case, except for pathological workload distributions [2,3] (Arturo Gonzalez Escribano). This limited loss is outweighed by the unlocked potential of automatic performance optimization, and, last but not least, the ultimate portability that is gained.

People collaborating

• Arjan J.C. van Gemund (Asstt Prof, EE Dept, Delft U. of Technology)
• Valentin Cardenoso Payo PhD (Assoc Prof, U. of Valladolid)
• Arturo Gonzalez Escribano (Asst Prof, U. of Valladolid)
• Frits Kuijlman (PhD student, CS Dept, Delft U. of Technology)

References

• [1] A.J.C. van Gemund, "The Importance of Synchronization Structure in Parallel Program Optimization," in Proc. 11th ACM Int'l Conf. on Supercomputing, pp. 164-171, Vienna, July 1997.
• [2] A. Gonzalez Escribano, V. Cardenoso, A.J.C. van Gemund, "On the Loss of Parallelism by Imposing Synchronization Structure," in Proc. 1st Int'l Conference on Parallel and Distributed Systems - Euro-PDS'97, Barcelona, June 1997.
• [3] A. Gonzalez Escribano, V. Cardenoso, A.J.C. van Gemund, "Approximating Unstructured Task Parallelism in Series-Parallel Form," in Proc. 3rd Annual Conference of the Advanced School for Computing and Imaging, Heijen, The Netherlands, pages 76-82, June 1997.
• [4] A.J.C. van Gemund, "Notes on SPC: A Parallel Programming Model," Technical Report 1-68340-44(1997)03, Delft University of Technology, Delft, The Netherlands, April 1997.
• [5] A.J.C. van Gemund, "SPC: A Model of Parallel Computation," Proceedings EuroPar'96, pages 397-400, Sept. 1996.