Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Parallel applications”

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.

At least 217 records · Page 12

DIstributed VIRtual System (DIVIRS) project

As outlined in our continuation proposal 92-ISI-50R (revised) on NASA cooperative agreement NCC2-539, we are (1) developing software, including a system manager and a job manager, that will manage available resources and that will enable programmers to develop and execute parallel applications in terms of a virtual configuration of processors, hiding the mapping to physical nodes; (2) developing communications routines that support the abstractions implemented in item one; (3) continuing the development of file and information systems based on the Virtual System Model; and (4) incorporating appropriate security measures to allow the mechanisms developed in items 1 through 3 to be used on an open network. The goal throughout our work is to provide a uniform model that can be applied to both parallel and distributed systems. We believe that multiprocessor systems should exist in the context of distributed systems, allowing them to be more easily shared by those that need them. Our work provides the mechanisms through which nodes on multiprocessors are allocated to jobs running within the distributed system and the mechanisms through which files needed by those jobs can be located and accessed.

Schorr, Herbert↗

Distributed-Memory Sparse Deep Neural Network Inference Using Global Arrays

Partitioned Global Address Space (PGAS) models exhibit tremendous promise in developing efficient and productive distributed-memory parallel applications. They have been used extensively in scientific computations due to conveniently offering a ``shared-memory''-like model and convenient interfaces that separate communication with synchronization. Traditionally, PGAS communication models have been applied to dense/contiguously distributed data, but most modern applications depict varied levels of sparsity. Existing PGAS models require certain adaptations to support distributed sparse computations, since associated computations often require matrix arithmetic, in addition to data movement. The Global Arrays toolkit from Pacific Northwest National Laboratory (PNNL) is one of the earliest PGAS models to combine one-sided data communication and distributed matrix operations and is still used in the popular NWChem quantum chemistry suite. Recently, we have expanded the Global Arrays toolkit to support common sparse operations, like sparse matrix-dense matrix multiplies (SpMM), sparse matrix-sparse matrix multiplication (SpGEMM) and Sampled Dense-Dense Matrix Multiplication (SDDMM). As it turns out, these operations are the bedrock of sparse Deep Learning (DL); sparse deep neural networks and Graph Neural Networks (GNNs) have gained increasing attention recently in achieving speedups on training and inference with reduced memory footprints. Unlike scientific applications in High Performance Computing (HPC), modern (distributed-memory capable) DL toolkits often rely on non-standardized and closed-source vendor software optimizations, creating challenges in software-hardware co-design at scale. Our goal is to support a variety of distributed-memory sparse matrix operations and helper functions in the newly created Sparse Global Arrays (SGA), such that it is possible to build portable and productive Machine Learning scenarios for algorithm/software and hardware codesign purposes. Contemporary data-parallel schemes for training/inference are undergoing a major overhaul since model replication limits scalability and causes resource inefficiencies. As such, we have adopted tensor parallelism in decomposing the model and inputs, to mitigate memory issues. Current implementation is built on top of MPI and uses CPUs to maximize the portability across the platforms.

Distributed computing, machine learning↗

NAS Experiences of Porting CM Fortran Codes to HPF on IBM SP2 and SGI Power Challenge

Current Connection Machine (CM) Fortran codes developed for the CM-2 and the CM-5 represent an important class of parallel applications. Several users have employed CM Fortran codes in production mode on the CM-2 and the CM-5 for the last five to six years, constituting a heavy investment in terms of cost and time. With Thinking Machines Corporation's decision to withdraw from the hardware business and with the decommissioning of many CM-2 and CM-5 machines, the best way to protect the substantial investment in CM Fortran codes is to port the codes to High Performance Fortran (HPF) on highly parallel systems. HPF is very similar to CM Fortran and thus represents a natural transition. Conversion issues involved in porting CM Fortran codes on the CM-5 to HPF are presented. In particular, the differences between data distribution directives and the CM Fortran Utility Routines Library, as well as the equivalent functionality in the HPF Library are discussed. Several CM Fortran codes (Cannon algorithm for matrix-matrix multiplication, Linear solver Ax=b, 1-D convolution for 2-D datasets, Laplace's Equation solver, and Direct Simulation Monte Carlo (DSMC) codes have been ported to Subset HPF on the IBM SP2 and the SGI Power Challenge. Speedup ratios versus number of processors for the Linear solver and DSMC code are presented.

Saini, Subhash↗

Performance Optimization Methods for a Memory-Bound, Unstructured-Grid CFD Application on Massively Parallel GPU Platforms

Computational performance of the FUN3D unstructured-grid computational fluid dynamics (CFD) application on massively parallel GPU environments is memory-bound and highly dependent upon efficient reads from and atomic updates to the irregular cell-, edge-, and node-based data structures. In this talk, we present recent efforts into optimizing select performance-critical kernels on NVIDIA Tesla V100 and A100 GPUs and AMD CDNA MI100 GPUs. A novel use of L2 cache residency controls and asynchronous loads into on-chip shared memory are explored on the A100 GPU for the sparse iterative solver, which is dominated by mixed-precision, sparse matrix vector multiplication. Demonstrations show that these methods improve global memory bandwidth utilization by 13.5% on the A100 GPU. Several techniques are also presented that use registers and/or shared memory to facilitate array transposition and aggregation which combine to reduce the frequency and increase the cache efficiency of floating-point atomic updates to the irregular data structures. These methods are demonstrated to improve the kernel throughput by nearly 500% on select kernels on the AMD MI100 over atomic updates directly to global memory. Overall, both V100 and A100 GPUs outperformed the MI100 GPU on kernels dominated by double-precision atomic updates; however, the techniques demonstrated here reduced the performance gap and improved the MI100 performance.

GPU CPU unstructured CFD memory↗

The Automated Instrumentation and Monitoring System (AIMS): Design and Architecture

Whether a researcher is designing the 'next parallel programming paradigm', another 'scalable multiprocessor' or investigating resource allocation algorithms for multiprocessors, a facility that enables parallel program execution to be captured and displayed is invaluable. Careful analysis of such information can help computer and software architects to capture, and therefore, exploit behavioral variations among/within various parallel programs to take advantage of specific hardware characteristics. A software tool-set that facilitates performance evaluation of parallel applications on multiprocessors has been put together at NASA Ames Research Center under the sponsorship of NASA's High Performance Computing and Communications Program over the past five years. The Automated Instrumentation and Monitoring Systematic has three major software components: a source code instrumentor which automatically inserts active event recorders into program source code before compilation; a run-time performance monitoring library which collects performance data; and a visualization tool-set which reconstructs program execution based on the data collected. Besides being used as a prototype for developing new techniques for instrumenting, monitoring and presenting parallel program execution, AIMS is also being incorporated into the run-time environments of various hardware testbeds to evaluate their impact on user productivity. Currently, the execution of FORTRAN and C programs on the Intel Paragon and PALM workstations can be automatically instrumented and monitored. Performance data thus collected can be displayed graphically on various workstations. The process of performance tuning with AIMS will be illustrated using various NAB Parallel Benchmarks. This report includes a description of the internal architecture of AIMS and a listing of the source code.

Yan, Jerry C.↗

Proactive Fault Tolerance Using Preemptive Migration

Proactive fault tolerance (FT) in high-performance computing is a concept that prevents compute node failures from impacting running parallel applications by preemptively migrating application parts away from nodes that are about to fail. This paper provides a foundation for proactive FT by defining its architecture and classifying implementation options. This paper further relates prior work to the presented architecture and classification, and discusses the challenges ahead for needed supporting technologies.

97 MATHEMATICS AND COMPUTING↗

A Simple XML Producer-Consumer Protocol

There are many different projects from government, academia, and industry that provide services for delivering events in distributed environments. The problem with these event services is that they are not general enough to support all uses and they speak different protocols so that they cannot interoperate. We require such interoperability when we, for example, wish to analyze the performance of an application in a distributed environment. Such an analysis might require performance information from the application, computer systems, networks, and scientific instruments. In this work we propose and evaluate a standard XML-based protocol for the transmission of events in distributed systems. One recent trend in government and academic research is the development and deployment of computational grids. Computational grids are large-scale distributed systems that typically consist of high-performance compute, storage, and networking resources. Examples of such computational grids are the DOE Science Grid, the NASA Information Power Grid (IPG), and the NSF Partnerships for Advanced Computing Infrastructure (PACIs). The major effort to deploy these grids is in the area of developing the software services to allow users to execute applications on these large and diverse sets of resources. These services include security, execution of remote applications, managing remote data, access to information about resources and services, and so on. There are several toolkits for providing these services such as Globus, Legion, and Condor. As part of these efforts to develop computational grids, the Global Grid Forum is working to standardize the protocols and APIs used by various grid services. This standardization will allow interoperability between the client and server software of the toolkits that are providing the grid services. The goal of the Performance Working Group of the Grid Forum is to standardize protocols and representations related to the storage and distribution of performance data. These standard protocols and representations must support tasks such as profiling parallel applications, monitoring the status of computers and networks, and monitoring the performance of services provided by a computational grid. This paper describes a proposed protocol and data representation for the exchange of events in a distributed system. The protocol exchanges messages formatted in XML and it can be layered atop any low-level communication protocol such as TCP or UDP Further, we describe Java and C++ implementations of this protocol and discuss their performance. The next section will provide some further background information. Section 3 describes the main communication patterns of our protocol. Section 4 describes how we represent events and related information using XML. Section 5 describes our protocol and Section 6 discusses the performance of two implementations of the protocol. Finally, an appendix provides the XML Schema definition of our protocol and event information.

Smith, Warren↗

Effective Boundary Treatment for the Biharmonic Dirichlet Problem

The biharmonic equation can be rewritten as a system of two Poisson equations. Multigrid solution of this system is expected to converge with the same amount of work as solving two Poisson equations, requiring less than 70 floating point operations (scalar multiply or addition) per fine grid point to reach a solution using an FMG algorithm. For periodic boundary conditions, this goal is attained by simple, straightforward application of multigrid. For Dirichlet boundary conditions, however, convergence is impeded by poor interaction with the boundaries. Attempts to overcome the slowness without specifically addressing the boundaries have resulted in multigrid algorithms not attaining the Poisson convergence rate. We present three methods of boundary treatment with which full multigrid efficiency can be obtained. All implement an approach described by Brandt, concentrating some additional effort near the boundary. The first approach simply adds a number of relaxation sweeps over points close to the boundary. The second uses joint relaxation on near-boundary points. The third method takes something from each of the first two methods, resulting in a solver more suitable for highly parallel applications.

Brandt, A.↗

Single Degree-of-Freedom Modeling of SLS Liquid Hydrogen Pre-Valve Flow Guide to Enable Rapid Transient Analysis

A unique single degree-of-freedom approximation technique has been developed to enable rapid application of a temporally-defined multi-spectral semi-narrow-band loading for generation of realistic stress/cycle values compared to a resonant analysis. The technique uses the harmonic analysis at resonance of a high-fidelity finite element model to produce a transfer function, which is then used to calibrate the response of the SDOF model. A standard numerical ordinary differential equation solver is then used to obtain the temporal response, and its histogram is used in a fatigue/fracture model. This technique is related to other SDOF methods used widely in industry, such as Miles' Equation and the Shock Response Spectra, but it is unique in that it produces a realistic time history of the response. The most obvious error in the process, which is the effect of closely-spaced modes, was also assessed using the parallel application of several SDOF models, and the error is shown to be small. The application of this unique and tractable reduced-order methodology has enabled the SLS program to avoid substantial cost and schedule penalties if a redesign or change of material were required. It has also enabled quick analysis of a number of other structures undergoing the same or similar excitation fields, and quick assessment when the excitation and structural configuration has been altered due to design changes in the system.

Brown, Andrew M.↗

Evaluation of Simple Causal Message Logging for Large-Scale Fault Tolerant HPC Systems

The era of petascale computing brought machines with hundreds of thousands of processors. The next generation of exascale supercomputers will make available clusters with millions of processors. In those machines, mean time between failures will range from a few minutes to few tens of minutes, making the crash of a processor the common case, instead of a rarity. Parallel applications running on those large machines will need to simultaneously survive crashes and maintain high productivity. To achieve that, fault tolerance techniques will have to go beyond checkpoint/restart, which requires all processors to roll back in case of a failure. Incorporating some form of message logging will provide a framework where only a subset of processors are rolled back after a crash. In this paper, we discuss why a simple causal message logging protocol seems a promising alternative to provide fault tolerance in large supercomputers. As opposed to pessimistic message logging, it has low latency overhead, especially in collective communication operations. Besides, it saves messages when more than one thread is running per processor. Finally, we demonstrate that a simple causal message logging protocol has a faster recovery and a low performance penalty when compared to checkpoint/restart. Running NAS Parallel Benchmarks (CG, MG and BT) on 1024 processors, simple causal message logging has a latency overhead below 5%.

97 MATHEMATICS AND COMPUTING↗

NWChem: scalable parallel computational chemistry

NWChem is a general purpose computational chemistry code specifically designed to run on distributed memory parallel computers. The core functionality of the code focuses on molecular dynamics, Hartree-Fock and density functional theory methods for both plane-wave basis sets as well as Gaussian basis sets, tensor contraction engine based coupled cluster capabilities and combined quantum mechanics/molecular mechanics descriptions. It was realized from the beginning that scalable implementations of these methods required a programming paradigm inherently different from what message passing approaches could offer. In response a global address space library, the Global Array Toolkit, was developed. The programming model it offers is based on using predominantly one-sided communication. This model underpins most of the functionality in NWChem and the power of it is exemplified by the fact that the code scales to tens of thousands of processors. In this paper the core capabilities of NWChem are described as well as their implementation to achieve an efficient computational chemistry code with high parallel scalability. NWChem is a modern, open source, computational chemistry code1 specifically designed for large scale parallel applications2. To meet the challenges of developing efficient, scalable and portable programs of this nature a particular code design was adopted. This code design involved two main features. First of all, the code is build up in a modular fashion so that a large variety of functionality can be integrated easily. Secondly, to facilitate writing complex parallel algorithms the Global Array toolkit was developed. This toolkit allows one to write parallel applications in a shared memory like approach, but offers additional mechanisms to exploit data locality to lower communication overheads. This framework has proven to be very successful in computational chemistry but is applicable to any engineering domain. Within the context created by the features above NWChem has grown into a general purpose computational chemistry code that supports a wide variety of energy expressions and capabilities to calculate properties based there upon. The main energy expressions are classical mechanics force fields, Hartree-Fock and DFT both for finite systems and condensed phase systems, coupled cluster, as well as QM/MM. For most energy expressions single point calculations, geometry optimizations, excited states, and other properties are available. Below we briefly discuss each of the main energy expressions and the critical points involved in scalable implementations thereof.

99 GENERAL AND MISCELLANEOUS↗

Optimizing Irregular Communication with Neighborhood Collectives and Locality-Aware Parallelism

Irregular communication often limits both the performance and scalability of parallel applications. Typically, applications individually implement irregular communication as point-to-point, and any optimizations are integrated directly into the application. As a result, these optimizations lack portability. It is difficult to optimize point-to-point messages within MPI, as the interface for single messages provides no information on the collection of all communication to be performed. However, the persistent neighbor collective API, released in the MPI 4 standard, provides an interface for portable optimizations of irregular communication within MPI libraries. This paper presents methods for implementing existing optimizations for irregular communication within neighborhood collectives, analyzes the impact of replacing point-to-point communication in existing codebases such as Hypre BoomerAMG with neighborhood collectives, and finally shows up to a 1.38x speedup on sparse matrix-vector multiplication communication within a BoomerAMG solve through the use of our optimized neighbor collectives. Here, the authors analyze three implementations of persistent neighborhood collectives for Alltoallv: an unoptimized wrapper of standard point-to-point communication, and two locality-aware aggregating methods. The second locality-aware implementation exposes an non-standard interface to perform additional optimization, and the authors present the additional 0.07x speedup from the extended interface. All optimizations are available in an open-source codebase, MPI Advance, which sits on top of MPI, allowing for optimizations to be added into existing codebases regardless of the system MPI install.

AMG↗

Achieving energy efficiency during collective communications

Energy consumption has become a major design constraint in modern computing systems. With the advent of petaflops architectures, power-efficient software stacks have become imperative for scalability. Techniques such as dynamic voltage and frequency scaling (called DVFS) and CPU clock modulation (called throttling) are often used to reduce the power consumption of the compute nodes. To avoid significant performance losses, these techniques should be used judiciously during parallel application execution. For example, its communication phases may be good candidates to apply the DVFS and CPU throttling without incurring a considerable performance loss. They are often considered as indivisible operations although little attention is being devoted to the energy saving potential of their algorithmic steps. In this work, two important collective communication operations, all-to-all and allgather, are investigated as to their augmentation with energy saving strategies on the per-call basis. The experiments prove the viability of such a fine-grain approach. They also validate a theoretical power consumption estimate for multicore nodes proposed here. While keeping the performance loss low, the obtained energy savings were always significantly higher than those achieved when DVFS or throttling were switched on across the entire application run

36 MATERIALS SCIENCE↗

Using Deep Learning for Automated Communication Pattern Characterization: Little Steps and Big Challenges

Characterization of a parallel application’s communication patterns can be useful for performance analysis, debugging, and system design. However, obtaining and interpreting a characterization can be difficult. AChax implements an approach that uses search and a library of known communication patterns to automatically characterize communication patterns. Our approach has some limitations that reduce its effectiveness for the patterns and pattern combinations used by some real-world applications. By viewing AChax’s pattern recognition problem as an image recognition problem, it may be possible to use deep learning to address these limitations. In this position paper, we present our current ideas regarding the benefits and challenges of integrating deep learning into AChax and our conclusion that a hybrid approach combining deep learning classification, regression, and the existing AChax approach may be the best long-term solution to the problem of parameterizing recognized communication patterns.

Roth, Philip↗

Improved Accuracy for Automated Communication Pattern Characterization Using Communication Graphs and Aggressive Search Space Pruning

An understanding of a parallel application’s communication behavior is useful for a range of activities including debugging and optimization, job scheduling, target system selection, and system design. Because it can be challenging to understand communication behavior, especially for those who lack expertise or who are not familiar with the application, I and two colleagues recently developed an automated, search-based approach for recognizing and parameterizing application communication behavior using a library of common communication patterns. This initial approach was effective for characterizing the behavior of many workloads, but I identified some combinations of communication patterns for which the method was inefficient or would fail. In this paper, I discuss one such troublesome pattern combination and propose modifications to the recognition method to handle it. Specifically, I propose an alternative approach that uses communication graphs instead of traditional communication matrices to improve recognition accuracy for collective communication operations, and that uses a non-greedy recognition technique to avoid search space dead-ends that trap the original greedy recognition approach. My modified approach uses aggressive search space pruning and heuristics to control the potential for state explosion caused by its non-greedy pattern recognition method. I demonstrate the improved recognition accuracy and pruning efficacy of the modified approach using several synthetic and real-world communication pattern combinations.

Roth, Philip↗

Designing Energy Efficient Communication Runtime Systems: A View from PGAS Models

As the march to the Exascale computing gains momentum, energy consumption of supercomputers has emerged to be the critical roadblock. While architectural innovations are imperative in achieving computing of this scale, it is largely dependent on the systems software to leverage the architectural innovations. Parallel applications in many computationally intensive domains have been designed to leverage these supercomputers, with legacy two-sided communication semantics using Message Passing Interface. At the same time, Partitioned Global Address Space Models are being designed which provide global address space abstractions and one-sided communication for exploiting data locality and communication optimizations. PGAS models rely on one-sided communication runtime systems for leveraging high-speed networks to achieve best possible performance. In this paper, we present a design for Power Aware One-Sided Communication Library - PASCoL. The proposed design detects communication slack, leverages Dynamic Voltage and Frequency Scaling (DVFS) and Interrupt driven execution to exploit the detected slack for energy efficiency. We implement our design and evaluate it using synthetic benchmarks for one-sided communication primitives, Put, Get and Accumulate and uniformly non-contiguous data transfers. Our performance evaluation indicates that we can achieve significant reduction in energy consumption without performance loss on multiple one-sided communication primitives. The achieved results are close to the theoretical peak available with the experimental test bed.

Vishnu, Abhinav↗

Establishing mesh topology in multi-material cells: enabling technology for robust and accurate multi-material simulations

Real world problems are typically multi-material, combining materials such as gases, liquids and solids that have very different properties. The material interfaces may be fixed in time or can be a part of the solution, as in fluid-structure interactions or air-water dynamics, and therefore move and change shape. In such problems the computational mesh may be non-conformal to interfaces due to complexity of these interfaces, presence of small fractions of materials, or because the mesh does not move with the flow, as in the arbitrary Lagrangian–Eulerian (ALE) methods. In order to solve problems of interest on such meshes, interface reconstruction methods are usually used to recover an approximation of material regions within the cells. For a cell intersecting multiple material regions, these approximations of contained subregions can be considered as single-material subcells in a local mesh that we call a minimesh. In this paper, we discuss some of the requirements that discretization methods have on topological information in the resulting hierarchical meshes and present an approach that allows incorporating the buildup of sufficiently detailed topology into the nested dissections based PLIC-type reconstruction algorithms (e.g. Volume-of-Fluid, Moment-of-Fluid) in an efficient and robust manner. Specifically, we describe the X-MOF interface reconstruction algorithm in 2D, which extends the Moment-Of-Fluid (MOF) method to include the topology of minimeshes created inside of multi-material cells and parent-child relations between corresponding mesh entities on different hierarchy levels. X-MOF retains the property of being local to a cell and not requiring external communication, which makes it suitable for massively parallel applications. Here, we demonstrate some scaling results for the X-MOF implementation in Tangram, a modern interface reconstruction framework for exascale computing.

97 MATHEMATICS AND COMPUTING↗