SEARCH · Search NASA
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
The programming tools and parallel algorithms created for the Massively Parallel Processor (MPP) located at the NASA Goddard Space Center are discussed. A user-friendly environment for high level language parallel algorithm development was developed. The issues involved in implementing certain algorithms on the MPP were researched. The expected results were compared with the actual results.
Most parallel scientific programs contain compiler directives (pragmas) such as those from OpenMP, explicit calls to runtime library procedures such as those implementing the Message Passing Interface (MPI), or compiler-specific language extensions such as those provided by CUDA. By contrast, the recent Fortran standards empower developers to express parallel algorithms without directly referencing lower-level parallel programming models. Fortran’s parallel features place the language within the Partitioned Global Address Space (PGAS) class of programming models. When writing programs that exploit data-parallelism, application developers often find it straightforward to develop custom parallel algorithms. Problems involving complex, heterogeneous, staged calculations, however, pose much greater challenges. Such applications require careful coordination of tasks in a manner that respects dependencies prescribed by a directed acyclic graph. When rolling one’s own solution proves difficult, extending a customizable framework becomes attractive. The paper presents the design, implementation, and use of the Framework for Extensible Asynchronous Task Scheduling (FEATS), which we believe to be the first task-scheduling tool written in modern Fortran. We describe the benefits and compromises associated with choosing Fortran as the implementation language, and we propose ways in which future Fortran standards can best support the use case in this paper.
This paper reports on the performance of five parallel algorithms for simulating a fully associative cache operating under the LRU (Least-Recently-Used) replacement policy. Three of the algorithms are SIMD, and are implemented on the MasPar MP-2 architecture. Two other algorithms are parallelizations of an efficient serial algorithm on the Intel Paragon. One SIMD algorithm is quite simple, but its cost is linear in the cache size. The two other SIMD algorithm are more complex, but have costs that are independent on the cache size. Both the second and third SIMD algorithms compute all stack distances; the second SIMD algorithm is completely general, whereas the third SIMD algorithm presumes and takes advantage of bounds on the range of reference tags. Both MIMD algorithm implemented on the Paragon are general and compute all stack distances; they differ in one step that may affect their respective scalability. We assess the strengths and weaknesses of these algorithms as a function of problem size and characteristics, and compare their performance on traces derived from execution of three SPEC benchmark programs.
This paper describes two models of the cost of data movement in parallel numerical alorithms. One model is a generalization of an approach due to Hockney, and is suitable for shared memory multiprocessors where each processor has vector capabilities. The other model is applicable to highly parallel nonshared memory MIMD systems. In this second model, algorithm performance is characterized in terms of the communication network design. Techniques used in VLSI complexity theory are also brought in, and algorithm-independent upper bounds on system performance are derived for several problems that are important to scientific computation.
The results of the implementation of a Navier-Stokes algorithm on three parallel/vector computers are presented. The object of this research is to determine how well, or poorly, a single numerical algorithm would map onto three different architectures. The algorithm is a compact difference scheme for the solution of the incompressible, two-dimensional, time-dependent Navier-Stokes equations. The computers were chosen so as to encompass a variety of architectures. They are the following: the MPP, an SIMD machine with 16K bit serial processors; Flex/32, an MIMD machine with 20 processors; and Cray/2. The implementation of the algorithm is discussed in relation to these architectures and measures of the performance on each machine are given. The basic comparison is among SIMD instruction parallelism on the MPP, MIMD process parallelism on the Flex/32, and vectorization of a serial code on the Cray/2. Simple performance models are used to describe the performance. These models highlight the bottlenecks and limiting factors for this algorithm on these architectures. Finally, conclusions are presented.
Two parallel algorithms for standard cell placement using simulated annealing are developed to run on distributed-memory message-passing hypercube multiprocessors. The cells can be mapped in a two-dimensional area of a chip onto processors in an n-dimensional hypercube in two ways, such that both small and large cell exchange and displacement moves can be applied. The computation of the cost function in parallel among all the processors in the hypercube is described, along with a distributed data structure that needs to be stored in the hypercube to support the parallel cost evaluation. A novel tree broadcasting strategy is used extensively for updating cell locations in the parallel environment. A dynamic parallel annealing schedule estimates the errors due to interacting parallel moves and adapts the rate of synchronization automatically. Two novel approaches in controlling error in parallel algorithms are described: heuristic cell coloring and adaptive sequence control.
The purpose of this project has been to identify and test suitable technologies for implementation of fluid flow solvers -- possibly coupled with structures and heat equation solvers -- on MIMD parallel computers. In the course of this investigation much attention has been paid to efficient domain decomposition strategies for ADI-type algorithms. Multi-partitioning derives its efficiency from the assignment of several blocks of grid points to each processor in the parallel computer. A coarse-grain parallelism is obtained, and a near-perfect load balance results. In uni-partitioning every processor receives responsibility for exactly one block of grid points instead of several. This necessitates fine-grain pipelined program execution in order to obtain a reasonable load balance. Although fine-grain parallelism is less desirable on many systems, especially high-latency networks of workstations, uni-partition methods are still in wide use in production codes for flow problems. Consequently, it remains important to achieve good efficiency with this technique that has essentially been superseded by multi-partitioning for parallel ADI-type algorithms. Another reason for the concentration on improving the performance of pipeline methods is their applicability in other types of flow solver kernels with stronger implied data dependence. Analytical expressions can be derived for the size of the dynamic load imbalance incurred in traditional pipelines. From these it can be determined what is the optimal first-processor retardation that leads to the shortest total completion time for the pipeline process. Theoretical predictions of pipeline performance with and without optimization match experimental observations on the iPSC/860 very well. Analysis of pipeline performance also highlights the effect of uncareful grid partitioning in flow solvers that employ pipeline algorithms. If grid blocks at boundaries are not at least as large in the wall-normal direction as those immediately adjacent to them, then the first processor in the pipeline will receive a computational load that is less than that of subsequent processors, magnifying the pipeline slowdown effect. Extra compensation is needed for grid boundary effects, even if all grid blocks are equally sized.
The main goal of this project was efficient distributed parallel and workstation cluster implementations of Newton-Krylov-Schwarz (NKS) solvers for implicit Computational Fluid Dynamics (CFD.) "Newton" refers to a quadratically convergent nonlinear iteration using gradient information based on the true residual, "Krylov" to an inner linear iteration that accesses the Jacobian matrix only through highly parallelizable sparse matrix-vector products, and "Schwarz" to a domain decomposition form of preconditioning the inner Krylov iterations with primarily neighbor-only exchange of data between the processors. Prior experience has established that Newton-Krylov methods are competitive solvers in the CFD context and that Krylov-Schwarz methods port well to distributed memory computers. The combination of the techniques into Newton-Krylov-Schwarz was implemented on 2D and 3D unstructured Euler codes on the parallel testbeds that used to be at LaRC and on several other parallel computers operated by other agencies or made available by the vendors. Early implementations were made directly in Massively Parallel Integration (MPI) with parallel solvers we adapted from legacy NASA codes and enhanced for full NKS functionality. Later implementations were made in the framework of the PETSC library from Argonne National Laboratory, which now includes pseudo-transient continuation Newton-Krylov-Schwarz solver capability (as a result of demands we made upon PETSC during our early porting experiences). A secondary project pursued with funding from this contract was parallel implicit solvers in acoustics, specifically in the Helmholtz formulation. A 2D acoustic inverse problem has been solved in parallel within the PETSC framework.
Most parallel algorithms for VLSI CAD proposed to date have one important drawback: they work efficiently only on machines that they were designed for. As a result, algorithms designed to date are dependent on the architecture for which they are developed and do not port easily to other parallel architectures. A new project under way to address this problem is described. A Portable object-oriented parallel environment for CAD algorithms (ProperCAD) is being developed. The objectives of this research are (1) to develop new parallel algorithms that run in a portable object-oriented environment (CAD algorithms using a general purpose platform for portable parallel programming called CARM is being developed and a C++ environment that is truly object-oriented and specialized for CAD applications is also being developed); and (2) to design the parallel algorithms around a good sequential algorithm with a well-defined parallel-sequential interface (permitting the parallel algorithm to benefit from future developments in sequential algorithms). One CAD application that has been implemented as part of the ProperCAD project, flat VLSI circuit extraction, is described. The algorithm, its implementation, and its performance on a range of parallel machines are discussed in detail. It currently runs on an Encore Multimax, a Sequent Symmetry, Intel iPSC/2 and i860 hypercubes, a NCUBE 2 hypercube, and a network of Sun Sparc workstations. Performance data for other applications that were developed are provided: namely test pattern generation for sequential circuits, parallel logic synthesis, and standard cell placement.
Algorithms for direct volume rendering on parallel and vector processors are investigated. Volumes are transformed efficiently on parallel processors by dividing the data into slices and beams of voxels. Equal sized sets of slices along one axis are distributed to processors. Parallelism is achieved at two levels. Because each slice can be transformed independently of others, processors transform their assigned slices with no communication, thus providing maximum possible parallelism at the first level. Within each slice, consecutive beams are incrementally transformed using coherency in the transformation computation. Also, coherency across slices can be exploited to further enhance performance. This coherency yields the second level of parallelism through the use of the vector processing or pipelining. Other ongoing efforts include investigations into image reconstruction techniques, load balancing strategies, and improving performance.
Algorithms for assembling in parallel the sparse system of linear equations that result from finite difference or finite element discretizations of elliptic partial differential equations, such as those that arise in structural engineering are developed. Parallel linear stationary iterative algorithms and parallel preconditioned conjugate gradient algorithms are developed for solving these systems. In addition, a model for comparing parallel algorithms on array architectures is developed and results of this model for the algorithms are given.
Parallel conjugate gradient algorithms for the computation of multibody dynamics are developed for the specialized case of a robot manipulator. For an n-dimensional positive-definite linear system, the Classical Conjugate Gradient (CCG) algorithms are guaranteed to converge in n iterations, each with a computation cost of O(n); this leads to a total computational cost of O(n sq) on a serial processor. A conjugate gradient algorithms is presented that provide greater efficiency using a preconditioner, which reduces the number of iterations required, and by exploiting parallelism, which reduces the cost of each iteration. Two Preconditioned Conjugate Gradient (PCG) algorithms are proposed which respectively use a diagonal and a tridiagonal matrix, composed of the diagonal and tridiagonal elements of the mass matrix, as preconditioners. Parallel algorithms are developed to compute the preconditioners and their inversions in O(log sub 2 n) steps using n processors. A parallel algorithm is also presented which, on the same architecture, achieves the computational time of O(log sub 2 n) for each iteration. Simulation results for a seven degree-of-freedom manipulator are presented. Variants of the proposed algorithms are also developed which can be efficiently implemented on the Robot Mathematics Processor (RMP).
Report presents parallel algorithms and special parallel architecture for computation of forward dynamics of robotics manipulators. Products of effort to find best method of parallel computation to achieve required computational efficiency. Significant speedup of computation anticipated as well as cost reduction.
Explore the source record for details and available documents.
A parallel algorithm for the solution of the generalized eigenproblem in linear elastic finite element analysis, (K)(phi)=(M)(phi)(omega), where (K) and (M) are of order N, and (omega) is of order q is presented. The parallel algorithm is based on a completely connected parallel architecture in which each processor is allowed to communicate with all other processors. The algorithm has been successfully implemented on a tightly coupled multiple-instruction-multiple-data (MIMD) parallel processing computer, Cray X-MP. A finite element model is divided into m domains each of which is assumed to process n elements. Each domain is then assigned to a processor, or to a logical processor (task) if the number of domains exceeds the number of physical processors. The macro-tasking library routines are used in mapping each domain to a user task. Computational speed-up and efficiency are used to determine the effectiveness of the algorithm. The effect of the number of domains, the number of degrees-of-freedom located along the global fronts and the dimension of the subspace on the performance of the algorithm are investigated. For a 64-element rectangular plate, speed-ups of 1.86, 3.13, 3.18 and 3.61 are achieved on two, four, six and eight processors, respectively.
New algorithm developed to overcome deficiency of skyline storage scheme by use of variable-band storage scheme. Exploits both parallel and vector capabilities of modern high-performance computers. Gives engineers and designers opportunity to include more design variables and constraints during optimization of structures. Enables use of more refined finite-element meshes to obtain improved understanding of complex behaviors of aerospace structures leading to better, safer designs. Not only attractive for current supercomputers but also for next generation of shared-memory supercomputers.
The papers contained in this volume provide an overview of the advances made in a number of aspects of computational mechanics, identify some of the anticipated industry needs in this area, discuss the opportunities provided by new hardware and parallel algorithms, and outline some of the current government programs in computational mechanics. Papers are included on advances and trends in parallel algorithms, supercomputers for engineering analysis, material modeling in nonlinear finite-element analysis, the Navier-Stokes computer, and future finite-element software systems.
An implementation of the alternating direction implicit (ADI) method for the Control Data STAR-100 computer is presented and analyzed. Two parallel algorithms, both of which are most efficient when used to solve many independent tridiagonal systems of equations, are discussed relative to their usefulness in an ADI implementation on the STAR-100 computer. It is shown that it may be desirable to alternate between the parallel algorithms as the direction of implicitness is alternated in order to eliminate the data rearrangement which would otherwise be required. The applicability of the two parallel tridiagonal solvers to several other numerical algorithms is also discussed.