|Distributed task pool implementation|
|Marco Danelutto - Dept. Computer Science - University of Pisa - Italy|
Lorenza Rovetti - Dept. Computer Science - University of Pisa - Italy
Parallel computations of independent tasks, coming either from a task stream or from data structure decomposition, can be efficiently implemented by exploiting macro data flow techniques. In this work we discuss a new technique that can be used to implement a distributed and efficient fireable macro data flow instruction repository (task pool). In particular, we discuss some experimental results that demonstrate that the implementation technique used to implement the distributed task pool is scalable and efficient. The results achieved will be used to improve existing prototypes that exploit macro data flow techniques in the execution of structured parallel programs. These prototypes currently use completely centralized fireable macro data flow instruction repository.