We propose an innovative approach to structured exploitation of both data and control parallelism. Our approach is based on the clear distinction between data and control parallelism exploitation mechanisms. This separation leads to a programming model where data and control parallelism exploitation is managed by means of independent, orthogonal mechanisms. By exploiting this orthogonal set of mechanisms, clear semantic transformation and verification tools can be developed. We show here a preliminary definition of the programming model, some sample skeleton applications and we discuss the basis for the development of a clear semantic framework that can be used to develop semantics preserving transformation rules as well as semantic based reasoning on parallel program properties.
A framework for orthogonal data and control parallelism exploitation
CAMPA, SONIA;DANELUTTO, MARCO
2004-01-01
Abstract
We propose an innovative approach to structured exploitation of both data and control parallelism. Our approach is based on the clear distinction between data and control parallelism exploitation mechanisms. This separation leads to a programming model where data and control parallelism exploitation is managed by means of independent, orthogonal mechanisms. By exploiting this orthogonal set of mechanisms, clear semantic transformation and verification tools can be developed. We show here a preliminary definition of the programming model, some sample skeleton applications and we discuss the basis for the development of a clear semantic framework that can be used to develop semantics preserving transformation rules as well as semantic based reasoning on parallel program properties.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.