SPMD Model Research Articles

Algorithms for Parallelizing a Mathematical Model of Forest Fires on Supercomputers and Theoretical Estimates for the Efficiency of Parallel Programs

The author develops algorithms of parallel implementation of a mathematical model of forest fires to model processes of their occurrence, development, and propagation taking into account physical and chemical processes. Approaches to parallel implementation and schemes of decomposition of solution domains for two-dimensional cases are proposed. Coarse-grained parallelization methods are proposed to be used in the SPMD model of calculations. Formulas are given that make it possible to theoretically estimate the efficiency of parallel programs.

An Extensible System for Multilevel Automatic Data Partition and Mapping

Automatic data distribution is a key feature to obtain efficient implementations from abstract and portable parallel codes. We present a highly efficient and extensible runtime library that integrates techniques for automatic data partition and mapping. It uses a novel approach to define an abstract interface and a plug-in system to encapsulate different types of regular and irregular techniques, helping to generate codes which are independent of the exact mapping functions selected. Currently, it supports hierarchical tiling of arrays with dense and stride domains, that allows the implementation of both data and task parallelism using a SPMD model. It automatically computes appropriate domain partitions for a selected virtual topology, mapping them to available processors with static or dynamic load-balancing techniques. Our library also allows the construction of reusable communication patterns that efficiently exploit MPI communication capabilities. The use of our library greatly reduces the complexity of data distribution and communication, hiding the details of the underlying architecture. The library can be used as an abstract layer for building generic tiling operations as well. Our experimental results show that the use of this library allows to achieve similar performance as carefully-implemented manual versions for several, well-known parallel kernels and benchmarks in distributed and multicore systems, and substantially reduces programming effort.

Domain Decomposition Approach for Automatic Parallel Generation of Tetrahedral Grids

AbstractThe desire to simulate even more geometrical and physical features of technical structures and the availability of parallel computers and parallel numerical solvers which can exploit the power of these machines have led to a steady increase in the number of the grid elements used. Memory requirements and computational time are too large for usual serial PCs. An a priori partitioning algorithm for the parallel generation of 3D non-overlapping compatible unstructured meshes based on a CAD surface description is presented in this paper. Emphasis is placed on practical issues and implementation rather than on theoretical complexity. In order to achieve robustness of the algorithm with respect to the geometrical shape of the structure, the authors propose that there should be several or many but relatively simple algorithmic steps. The geometrical domain decomposition approach has been applied. It allows us to use standard 2D and 3D high-quality Delaunay mesh generators for independent and simultaneous volume meshing. Different aspects of load balancing methods are also explored in the paper. The MPI library and SPMD model are used for parallel grid generator implementation. Several 3D examples are given.

Abstraction and performance in the design of parallel programs: an overview of the SAT approach

SAT stands for Stages And Tranformations and is the name of an approach to the high-level, performance-directed design of parallel programs. The target programs obtained with this approach are sequences of internally parallel stages, i.e., they fall within the SPMD model. Formal program transformations are used for deriving each parallel stage and optimizing the combination of several stages. The main advantage of the SAT approach is the comparatively high level of abstraction at which performance-relevant design decisions can be made. One consequence is that choices become much clearer than at the code level, enhancing comparability; another is that there is an increased potential for automation of the programming process. This paper summarizes the approach and assesses its usefulness, with emphasis on one basic parallel programming pattern, the list homomorphism, which captures the divide-and-conquer paradigm. We survey the transformational theory, provide a range of practical examples, and discuss the potential for automation and also the demands made on application programmers and implementers.

Titanium: a high-performance Java dialect

Titanium is a language and system for high-performance parallel scientific computing. Titanium uses Java as its base, thereby leveraging the advantages of that language and allowing us to focus attention on parallel computing issues. The main additions to Java are immutable classes, multidimensional arrays, an explicitly parallel SPMD model of computation with a global address space, and zone-based memory management. We discuss these features and our design approach, and report progress on the development of Titanium, including our current driving application: a three-dimensional adaptive mesh refinement parallel Poisson solver. © 1998 John Wiley & Sons, Ltd.

Synchronization Minimization in a SPMD Execution Model

This paper presents an algorithm for synchronization placement when using a SPMD execution model, where synchronizations are enforced only when there exists a cross-processor data dependence. In this paper, we investigate two scheduling techniques, loop-based and data-based, both of which use a SPMD model. Using scheduling information from previous stages in the compilation process, a new technique to determine potential cross-processor data dependences is presented. Given the minimum number of cross-processor data dependences that must be satisfied, a new optimization is used so as to minimize the number of synchronization points needed to satisfy them. This algorithm has been successfully implemented in an experimental compiler. Initial experimental data show this technique to be very effective, outperforming existing methods.

Shared variable oriented parallel precompiler for SPMD model

For the moment, commercial parallel computer systems with distributed memory architecture are usually provided with parallel FORTRAN or parallel C compliers, which are just traditional sequential FORTRAN or C compilers expanded with communication statements. Programmers suffer from writing parallel programs with communication statements.

Bounded quantifications for iteration and concurrency in logic programming

The only means for repetition in most logic programming languages, including Prolog, is recursion. Definite iteration is introduced in logic programming languages through the bounded quantification construct. Firstly, it is claimed that this construct is often, though not always, more natural than recursion for expressing relations that involve repetition. In particular, programs involving arrays and similar data structures are significantly simplified. Secondly, it is argued that bounded quantifications should be efficiently implementable on sequential computers and have a high potential for running in parallel, particularly on computers supporting the SPMD model of computation.