Algorithms exist for compiling Fortran D for MIMD distributed-memory machines, but are significantly restricted in the presence of procedure calls. This paper presents interprocedural analysis, optimization, and code generation algorithms for Fortran D that limit compilation to only one pass over each procedure. This is accomplished by collecting summary information after edits, then compiling procedures in reverse topological order to propagate necessary information. Delaying instantiation of the computation partition, communication, and dynamic data decomposition is key to enabling interprocedural optimization. Recompilation analysis preserves the benefits of separate compilation. Empirical results show that interprocedural optimization is crucial in achieving acceptable performance for a common application.

This paper presents Fortran D, a version of Fortran enhanced with data decomposition specifications. It is designed to support two fundamental stages of writing a data-parallel program: problem mapping using sophisticated array alignments, and machine mapping through a rich set of data distribution functions. We believe that Fortran D provides a simple machine-independent programming model for most numerical computations. We intend to evaluate its usefulness for both programmers and advanced compilers on a variety of parallel architectures.

An important problem facing numerous research projects on parallelizing compilers for distributed memory machines is that of automatically determining a suitable data partitioning scheme for a program. Most of the current projects leave this tedious problem almost entirely to the user. In this paper, we present a novel approach to the problem of automatic data partitioning. We introduce the notion of constraints on data distribution, and show how, based on performance considerations, a compiler identifies constraints to be imposed on the distribution of various data structures. These constraints are then combined by the compiler to obtain a complete and consistent picture of the data distribution scheme, one that offers good performance in terms of the overall execution time. We present results of a study we performed on Fortran programs taken from the Linpack and Eispack libraries and the Perfect Benchmarks to determine the applicability of our approach to real programs. The results a...

this article, we describe an execution model for supporting programs that use pointer-based dynamic data structures. This model uses a simple mechanism for migrating a thread of control based on the layout of heap-allocated data and introduces parallelism using a technique based on futures and lazy task creation. We intend to exploit this execution model using compiler analyses and automatic parallelization techniques. We have implemented a prototype system, which we call Olden, that runs on the Intel iPSC/860 and the Thinking Machines CM-5. We discuss our implementation and report on experiments with five benchmarks.

We present a simple and novel framework for generating blocked codes for high-performance machines with a memory hierarchy. Unlike traditional compiler techniques like tiling, which are based on reasoning about the control flow of programs, our techniques are based on reasoning directly about the flow of data through the memory hierarchy. Our data-centric transformations permit a more direct solution to the problem of enhancing data locality than current control-centric techniques do, and generalize easily to multiple levels of memory hierarchy. We buttress these claims with performance numbers for standard benchmarks from the problem domain of dense numerical linear algebra. The simplicity and intuitive appeal of our approach should make it attractive to compiler writers as well as to library writers. 1 Introduction Data reuse is imperative for good performance on modern high-performance computers because the memory architecture of these machines is a hierarchy in which the cost of ...

Array privatization is one of the most effective transformations for the exploitation of parallelism. In this paper, we present a technique for automatic array privatization. Our algorithm uses data flow analysis of array references to identify privatizable arrays intraprocedurally as well as interprocedurally. It employs static and dynamic resolution to determine the last value of a lived private array.We compare the result of automatic array privatization with that of manual array privatization and identify directions for future improvement. To enhance the effectiveness of our algorithm, wedevelop a goal directly technique to analysis symbolic variables in the present of conditional statements, loops and index arrays.

This paper addresses the problem of compiling perfectly nested loops for multicomputers (distributed memory machines). The relatively high communication startup costs in these machines renders frequent communication very expensive. Motivated by this, we present a method of aggregating a number of loop iterations into tiles where the tiles execute atomically -- a processor executing the iterations belonging to a tile receives all the data it needs before executing any one of the iterations in the tile, executes all the iterations in the tile and then sends the data needed by other processors. Since synchronization is not allowed during the execution of a tile, partitioning the iteration space into tiles must not result in deadlock. We first show the equivalence between the problem of finding partitions and the problem of determining the cone for a given set of dependence vectors. We then present an approach to partitioning the iteration space into deadlock-free tiles so that communicati...

Massively parallel MIMD distributed-memory machines can provide enormous computation power. However, the difficulty of developing parallel programs for these machines has limited their accessibility. This paper presents compiler algorithms to automatically derive efficient message-passing programs based on data decompositions. Optimizations are presented to minimize load imbalance and communication costs for both loosely synchronous and pipelined loops. These techniques are employed in the compiler being developed at Rice University for Fortran D, a version of Fortran enhanced with data decomposition specifications. 1 Introduction It is widely recognized that parallel computing represents the only plausible way to continue to increase the computational power available to computational scientists and engineers. However, parallel computers are not likely to be widely successful until they are easy to program. A major component in the success of vector supercomputers is the ability of ...

This paper discusses some of the issues involved in implementing a shared-address space programming model on large-scale, distributed-memory multiprocessors. While such a programming model can be implemented on both shared-memory and messagepassing architectures, we argue that the transparent, coherent caching of global data provided by many shared-memory architectures is of crucial importance. Because message-passing mechanisms are much more efficient than shared-memory loads and stores for certain types of interprocessor communication and synchronization operations, however, we argue for building multiprocessors that efficiently support both shared-memory and message-passing mechanisms. We describe an architecture, Alewife, that integrates support for shared-memory and message-passing through a simple interface; we expect the compiler and runtime system to cooperate in using appropriate hardware mechanisms that are most efficient for specific operations. We report on both integrated and exclusively shared-memory implementations of our runtime system and two applications. The integrated runtime system drastically cuts down the cost of communication incurred by the scheduling, load balancing, and certain synchronization operations. We also present preliminary performance results comparing the two systems.

This paper addresses the problem of partitioning data for distributed memory machines (multicomputers). In current day multicomputers, interprocessor communication is more time-consuming than instruction execution. If insufficient attention is paid to the data allocation problem, then the amount of time spent in interprocessor communication might be so high as to seriously undermine the benefits of parallelism. It is therefore worthwhile for a compiler to analyze patterns of data usage to determine allocation, in order to minimize interprocessor communication. We present a machineindependent analysis of communication-free partitions. We present a matrix notation to describe array accesses in fully parallel loops which lets us derive sufficient conditions for communication-free partitioning (decomposition) of arrays. In the case of a commonly occurring class of accesses, we present a problem formulation to minimize communication costs, when communication-free partitioning of arrays is not possible.