Compiling High Level Constructs to Distributed Memory Architectures

Compiling High Performance Fortran for Distributed-memory Systems

ming language for writing parallel programs. It is based on the Fortran 90 language, with extensions that enable the programmer to specify how array operations can be divided among multiple processors for increased performance. In HPF, the program specifies only the pattern in which the data is divided among the processors; the compiler automates the low-level details of synchronization and com...

Compiling for Distributed Memory Architectures

Using High-Level Performance Prediction in Compiling for Distributed Systems

In cost-driven program optimization performance feedback is either based on a model of the algorithm or on a model of the actually generated machine code. Especially in the case of a distributed-memory system, the di erence in abstraction is large. In this paper we study the trade-o between prediction at high (program) level and low (machine) level in the context of automatic optimization for m...

Automatic Distributed-Memory Parallelization and Code Generation using the Polyhedral Framework

Compiling for distributed-memory parallel architectures is considered very challenging. In spite of the large amount of work done to address this problem, no practical and efficient solution currently exists. We present new techniques for compilation of regular sequential programs for distributed-memory parallel architectures – typically, a cluster of multicores interconnected with a high perfo...

Automatic Vectorization of Communicationsfor

Optimizing communication is a key issue in compiling data-parallel languages for distributed memory architectures. We examine here the case of cyclic distribution, and we derive symbolic expressions for communication sets under the only assumption that the initial parallel loop is deened by aane expressions of the indices. This technique relys on unimodular changes of basis. Analysis of the pro...