Aroma: Language Support for Distributed Objects

accesses in the case of regular data structures. This is Aroma simplifies the task of parallelizing large achieved through aggregate objects that are distributed applications on multicomputers by providing applications over multiple computing nodes using directives similar to with a shared object space. Aroma supports both those used by parallelizing compilers. For data parallel traditional monolithic objects and aggregate object that computations, aggregate objects support the distribution can be partitioned across multiple nodes. Aggregate of data and associated computation, resulting in mostly objects support data parallelism efficiently. An Aroma local data accesses and good load balancing. Aroma data program consists of tasks that operate on shared objects. objects can also have synchronization properties Tasks typically execute on the node on which their input associated with them. This makes it possible to write data is located, thus minimizing communication. Shared applications without explicit locks or condition variables data objects have synchronization properties associated [8] if the synchronization and data consistency with them, making it possible to parallelize a large class requirements are simple, as is often the case. Knowing of applications without using explicit locks and condition the synchronization and access properties of objects also variables. In this paper, we present and justify the Aroma allows an efficient implementation. language features, and give examples of Aroma programs.