Data Parallel Language Extensions for Exploiting Locality in Irregular Problems

Many large-scale computational applications contain irregular data access patterns related to unstructured problem domains. Examples include nite element methods, computational uid dynamics, and molecular dynamics codes. Such codes are diicult to parallelize ef-ciently with current HPF compilers. However, most of these problems exhibit spatial locality. This property is exploited by our approach. In the sequential program, unstructured domains are accessed via in-direction arrays. We introduce a new directive that serves to identify indirection arrays and the boundaries of the associated domains. The data domains are distributed using Multiple Recursive Decomposition (MRD), a pseudo-regular distribution, which combines eecient implementation with good load balancing and communication behavior. Indi-rection arrays are aligned with the data arrays. Using the information provided in the directive, the compiler can produce a target program with signiicantly better performance than an approach based on indirect distributions and the inspector/executor paradigm.