Search NASA⌕ Search

SEARCH · Search NASA

Results for “distributed memory”

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 271 records · Page 15

A Study of Clustering Techniques and Hierarchical Matrix Formats for Kernel Ridge Regression

We present memory-efficient and scalable algorithms for kernel methods used in machine learning. Using hierarchical matrix approximations for the kernel matrix the memory requirements, the number of floating point operations, and the execution time are drastically reduced compared to standard dense linear algebra routines. We consider both the general H matrix hierarchical format as well as Hierarchically Semi-Separable (HSS) matrices. Furthermore, we investigate the impact of several preprocessing and clustering techniques on the hierarchical matrix compression. Effective clustering of the input leads to a ten-fold increase in efficiency of the compression. The algorithms are implemented using the STRUMPACK solver library. These results confirm that - with correct tuning of the hyperparameters - classification using kernel ridge regression with the compressed matrix does not lose prediction accuracy compared to the exact - not compressed - kernel matrix and that our approach can be extended to O(1M) datasets, for which computation with the full kernel matrix becomes prohibitively expensive. We present numerical experiments in a distributed memory environment up to 1,024 processors of the NERSC's Cori supercomputer using well-known datasets to the machine learning community that range from dimension 8 up to 784.

97 MATHEMATICS AND COMPUTING↗

Unit commitment with wind power generation: integrating wind forecast uncertainty and stochastic programming.

We present a computational framework for integrating the state-of-the-art Weather Research and Forecasting (WRF) model in stochastic unit commitment/energy dispatch formulations that account for wind power uncertainty. We first enhance the WRF model with adjoint sensitivity analysis capabilities and a sampling technique implemented in a distributed-memory parallel computing architecture. We use these capabilities through an ensemble approach to model the uncertainty of the forecast errors. The wind power realizations are exploited through a closed-loop stochastic unit commitment/energy dispatch formulation. We discuss computational issues arising in the implementation of the framework. In addition, we validate the framework using real wind speed data obtained from a set of meteorological stations. We also build a simulated power system to demonstrate the developments.

17 WIND ENERGY↗

Concurrent Collections (CnC): A new approach to parallel programming

A common approach in designing parallel languages is to provide some high level handles to manipulate the use of the parallel platform. This exposes some aspects of the target platform, for example, shared vs. distributed memory. It may expose some but not all types of parallelism, for example, data parallelism but not task parallelism. This approach must find a balance between the desire to provide a simple view for the domain expert and provide sufficient power for tuning. This is hard for any given architecture and harder if the language is to apply to a range of architectures. Either simplicity or power is lost. Instead of viewing the language design problem as one of providing the programmer with high level handles, we view the problem as one of designing an interface. On one side of this interface is the programmer (domain expert) who knows the application but needs no knowledge of any aspects of the platform. On the other side of the interface is the performance expert (programmer or program) who demands maximal flexibility for optimizing the mapping to a wide range of target platforms (parallel / serial, shared / distributed, homogeneous / heterogeneous, etc.) but needs no knowledge of the domain. Concurrent Collections (CnC) is based on this separation of concerns. The talk will present CnC and its benefits. About the speaker. Kathleen Knobe has focused throughout her career on parallelism especially compiler technology, runtime system design and language design. She worked at Compass (aka Massachusetts Computer Associates) from 1980 to 1991 designing compilers for a wide range of parallel platforms for Thinking Machines, MasPar, Alliant, Numerix, and several government projects. In 1991 she decided to finish her education. After graduating from MIT in 1997, she joined Digital Equipment’s Cambridge Research Lab (CRL). She stayed through the DEC/Compaq/HP mergers and when CRL was acquired and absorbed by Intel. She currently works in the Software and Services Group / Technology Pathfinding and Innovation.

97 MATHEMATICS AND COMPUTING↗

SPARC: Demonstrate burst-buffer-based checkpoint/restart on ATS-1.

Recent high-performance computing (HPC) platforms such as the Trinity Advanced Technology System (ATS-1) feature burst buffer resources that can have a dramatic impact on an application’s I/O performance. While these non-volatile memory (NVM) resources provide a new tier in the storage hierarchy, developers must find the right way to incorporate the technology into their applications in order to reap the benefits. Similar to other laboratories, Sandia is actively investigating ways in which these resources can be incorporated into our existing libraries and workflows without burdening our application developers with excessive, platform-specific details. This FY18Q1 milestone summaries our progress in adapting the Sandia Parallel Aerodynamics and Reentry Code (SPARC) in Sandia’s ATDM program to leverage Trinity’s burst buffers for checkpoint/restart operations. We investigated four different approaches with varying tradeoffs in this work: (1) simply updating job script to use stage-in/stage out burst buffer directives, (2) modifying SPARC to use LANL’s hierarchical I/O (HIO) library to store/retrieve checkpoints, (3) updating Sandia’s IOSS library to incorporate the burst buffer in all meshing I/O operations, and (4) modifying SPARC to use our Kelpie distributed memory library to store/retrieve checkpoints. Team members were successful in generating initial implementation for all four approaches, but were unable to obtain performance numbers in time for this report (reasons: initial problem sizes were not large enough to stress I/O, and SPARC refactor will require changes to our code). When we presented our work to the SPARC team, they expressed the most interest in the second and third approaches. The HIO work was favored because it is lightweight, unobtrusive, and should be portable to ATS-2. The IOSS work is seen as a long-term solution, and is favored because all I/O work (including checkpoints) can be deferred to a single library.

97 MATHEMATICS AND COMPUTING↗

An Adaptive Memory Interface Controller for Improving Bandwidth Utilization of Hybrid and Reconfigurable Systems

Emerging applications such as data mining, bioinformatics, knowledge discovery, social network analysis are irregular. They use data structures based on pointers or linked lists, such as graphs, unbalanced trees or unstructures grids, which generates unpredictable memory accesses. These data structures usually are large, but difficult to partition. These applications mostly are memory bandwidth bounded and have high synchronization intensity. However, they also have large amounts of inherent dynamic parallelism, because they potentially perform a task for each one of the element they are exploring. Several efforts are looking at accelerating these applications on hybrid architectures, which integrate general purpose processors with reconfigurable devices. Some solutions, which demonstrated significant speedups, include custom-hand tuned accelerators or even full processor architectures on the reconfigurable logic. In this paper we present an approach for the automatic synthesis of accelerators from C, targeted at irregular applications. In contrast to typical High Level Synthesis paradigms, which construct a centralized Finite State Machine, our approach generates dynamically scheduled hardware components. While parallelism exploitation in typical HLS-generated accelerators is usually bound within a single execution flow, our solution allows concurrently running multiple execution flow, thus also exploiting the coarser grain task parallelism of irregular applications. Our approach supports multiple, multi-ported and distributed memories, and atomic memory operations. Its main objective is parallelizing as many memory operations as possible, independently from their execution time, to maximize the memory bandwidth utilization. This significantly differs from current HLS flows, which usually consider a single memory port and require precise scheduling of memory operations. A key innovation of our approach is the generation of a memory interface controller, which dynamically maps concurrent memory accesses to multiple ports. We present a case study on a typical irregular kernel, Graph Breadth First search (BFS), exploring different tradeoffs in terms of parallelism and number of memories.

BFS↗

Scalable In Situ Computation of Lagrangian Representations via Local Flow Maps

In situ computation of Lagrangian flow maps to enable post hoc time-varying vector field analysis has recently become an active area of research. However, the current literature is largely limited to theoretical settings and lacks a solution to address scalability of the technique in distributed memory. To improve scalability, we propose and evaluate the benefits and limitations of a simple, yet novel, performance optimization. Our proposed optimization is a communication-free model resulting in local Lagrangian flow maps, requiring no message passing or synchronization between processes, intrinsically improving scalability, and thereby reducing overall execution time and alleviating the encumbrance placed on simulation codes from communication overheads. To evaluate our approach, we computed Lagrangian flow maps for four time-varying simulation vector fields and investigated how execution time and reconstruction accuracy are impacted by the number of GPUs per compute node, the total number of compute nodes, particles per rank, and storage intervals. Our study consisted of experiments computing Lagrangian flow maps with up to 67M particle trajectories over 500 cycles and used as many as 2048 GPUs across 512 compute nodes. In all, our study contributes an evaluation of a communication-free model as well as a scalability study of computing distributed Lagrangian flow maps at scale using in situ infrastructure on a modern supercomputer.

Sane, Sudhanshu↗

nu-TRLan User Guide Version 1.0: A High-Performance Software Package for Large-Scale Harmitian Eigenvalue Problems

The original software package TRLan, [TRLan User Guide], page 24, implements the thick restart Lanczos method, [Wu and Simon 2001], page 24, for computing eigenvalues {lambda} and their corresponding eigenvectors v of a symmetric matrix A: Av = {lambda}v. Its effectiveness in computing the exterior eigenvalues of a large matrix has been demonstrated, [LBNL-42982], page 24. However, its performance strongly depends on the user-specified dimension of a projection subspace. If the dimension is too small, TRLan suffers from slow convergence. If it is too large, the computational and memory costs become expensive. Therefore, to balance the solution convergence and costs, users must select an appropriate subspace dimension for each eigenvalue problem at hand. To free users from this difficult task, nu-TRLan, [LNBL-1059E], page 23, adjusts the subspace dimension at every restart such that optimal performance in solving the eigenvalue problem is automatically obtained. This document provides a user guide to the nu-TRLan software package. The original TRLan software package was implemented in Fortran 90 to solve symmetric eigenvalue problems using static projection subspace dimensions. nu-TRLan was developed in C and extended to solve Hermitian eigenvalue problems. It can be invoked using either a static or an adaptive subspace dimension. In order to simplify its use for TRLan users, nu-TRLan has interfaces and features similar to those of TRLan: (1) Solver parameters are stored in a single data structure called trl-info, Chapter 4 [trl-info structure], page 7. (2) Most of the numerical computations are performed by BLAS, [BLAS], page 23, and LAPACK, [LAPACK], page 23, subroutines, which allow nu-TRLan to achieve optimized performance across a wide range of platforms. (3) To solve eigenvalue problems on distributed memory systems, the message passing interface (MPI), [MPI forum], page 23, is used. The rest of this document is organized as follows. In Chapter 2 [Installation], page 2, we provide an installation guide of the nu-TRLan software package. In Chapter 3 [Example], page 3, we present a simple nu-TRLan example program. In Chapter 4 [trl-info structure], page 7, and Chapter 5 [trlan subroutine], page 14, we describe the solver parameters and interfaces in detail. In Chapter 6 [Solver parameters], page 21, we discuss the selection of the user-specified parameters. In Chapter 7 [Contact information], page 22, we give the acknowledgements and contact information of the authors. In Chapter 8 [References], page 23, we list reference to related works.

97 MATHEMATICS AND COMPUTING↗

Petascale direct numerical simulation of blood flow on 200K cores and heterogeneous architectures

We present a fast, petaflop-scalable algorithm for Stokesian particulate flows. Our goal is the direct simulation of blood, which we model as a mixture of a Stokesian fluid (plasma) and red blood cells (RBCs). Directly simulating blood is a challenging multiscale, multiphysics problem. We report simulations with up to 260 million deformable RBCs. The largest simulation amounts to 90 billion unknowns in space. In terms of the number of cells, we improve the state-of-the art by several orders of magnitude: the previous largest simulation, at the same physical fidelity as ours, resolved the flow of O(1,000-10,000) RBCs. Our approach has three distinct characteristics: (1) we faithfully represent the physics of RBCs by using nonlinear solid mechanics to capture the deformations of each cell; (2) we accurately resolve the long-range, N-body, hydrodynamic interactions between RBCs (which are caused by the surrounding plasma); and (3) we allow for highly non-uniform spatial distributions of RBCs. The new method has been implemented in the software library MOBO (for 'Moving Boundaries'). We designed MOBO to support parallelism at all levels, including inter-node distributed memory parallelism, intra-node shared memory parallelism, data parallelism (vectorization), and fine-grained multithreading for GPUs. We have implemented and optimized the majority of the computation kernels on both Intel/AMD x86 and NVidia's Tesla/Fermi platforms for single and double floating point precision. Overall, the code has scaled on 256 CPU-GPUs on the Teragrid's Lincoln cluster and on 200,000 AMD cores of the Oak Ridge National Laboratory's Jaguar PF system. In our largest simulation, we have achieved 0.7 Petaflops/s of sustained performance on Jaguar.

59 BASIC BIOLOGICAL SCIENCES↗

Parallel Distributed Hydrology Soil Vegetation Model (DHSVM) using global arrays

The Distributed Hydrology Soil Vegetation Model (DHSVM) code was parallelized for distributed memory computers using Global Arrays (GA). To analyze parallel performance, DHSVM was used to simulate two significantly sized river basins located in northwest continental United States and southwest Canada at 90~m resolution: (1) Clearwater (25,000~km2) and (2) Columbia (668,000~km2). Meteorological forcing applied to both basins was dynamically down-scaled from a global circulation model using the Weather Research and Forecasting (WRF) model and read into DHSVM as 2D maps each time step. We report parallel code speed up was significant. Run times for 1-year simulations were reduced by an order of magnitude for both test basins. A maximum speed up of 105 was attained with 480 processors while simulating the Columbia basin. Speed up was limited by input dominated tasks, particularly the input of meteorological forcing data.

54 ENVIRONMENTAL SCIENCES↗

Clock Synchronization in High-end Computing Environments: A Strategy for Minimizing Clock Variance at Runtime

We present a new software-based clock synchronization scheme that provides high precision time agreement among distributed memory nodes. The technique is designed to minimize variance from a reference chimer during runtime and with minimal time-request latency. Our scheme permits initial unbounded variations in time and corrects both slow and fast chimers (clock skew). An implementation developed within the context of the MPI message passing interface is described, and time coordination measurements are presented. Among our results, the mean time variance for a set of nodes improved from 20.0 milliseconds under standard Network Time Protocol (NTP) down to 2.29 secs under our scheme.

clock synchronization↗

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↗

Hierarchical computation in the SPMD programming model

Large-scale parallel machines are programmed mainly with the single program, multiple data (SPMD) model of parallelism. While this model has advantages of scalability and simplicity, it does not fit well with divide-and-conquer parallelism or hierarchical machines that mix shared and distributed memory. In this paper, we define the recursive single program, multiple data model (RSPMD) that extends SPMD with a hierarchical team mechanism to support hierarchical algorithms and machines. We implement this model in the Titanium language and describe how to eliminate a class of deadlocks by ensuring alignment of collective operations. We present application case studies evaluating the RSPMD model, showing that it enables divide-and-conquer algorithms such as sorting to be elegantly expressed and that team collective operations increase performance of conjugate gradient by up to a factor of two. The model also facilitates optimizations for hierarchical machines, improving scalability of particle in cell by 8x and performance of sorting and a stencil code by up to 40% and 14%, respectively.

Kamil, A↗

Integrating Asynchronous Task Parallelism with OpenSHMEM, In: OpenSHMEM 2016: OpenSHMEM and Related Technologies. Enhancing OpenSHMEM for Hybrid Environments

Partitioned Global Address Space (PGAS) programming models combine shared and distributed memory features, and provide a foundation for high-productivity parallel programming using lightweight one-sided communications. The OpenSHMEM programming interface has recently begun gaining popularity as a lightweight library-based approach for developing PGAS applications, in part through its use of a symmetric heap to realize more efficient implementations of global pointers than in other PGAS systems. However, current approaches to hybrid inter-node and intra-node parallel programming in OpenSHMEM rely on the use of multithreaded programming models (e.g., pthreads, OpenMP) that harness intra-node parallelism but are opaque to the OpenSHMEM runtime. This OpenSHMEM+X approach can encounter performance challenges such as bottlenecks on shared resources, long pause times due to load imbalances, and poor data locality. Furthermore, OpenSHMEM+X requires the expertise of hero-level programmers, compared to the use of just OpenSHMEM. All of these are hard challenges to mitigate with incremental changes. This situation will worsen as computing nodes increase their use of accelerators and heterogeneous memories.

Computer Science↗

On the Performance of an Algebraic Multigrid Solver on Multicore Clusters

Algebraic multigrid (AMG) solvers have proven to be extremely efficient on distributed-memory architectures. However, when executed on modern multicore cluster architectures, we face new challenges that can significantly harm AMG's performance. We discuss our experiences on such an architecture and present a set of techniques that help users to overcome the associated problems, including thread and process pinning and correct memory associations. We have implemented most of the techniques in a MultiCore SUPport library (MCSup), which helps to map OpenMP applications to multicore machines. We present results using both an MPI-only and a hybrid MPI/OpenMP model.

99 GENERAL AND MISCELLANEOUS↗

High-performance Computing Applied to Semantic Databases

To-date, the application of high-performance computing resources to Semantic Web data has largely focused on commodity hardware and distributed memory platforms. In this paper we make the case that more specialized hardware can offer superior scaling and close to an order of magnitude improvement in performance. In particular we examine the Cray XMT. Its key characteristics, a large, global shared-memory, and processors with a memory-latency tolerant design, offer an environment conducive to programming for the Semantic Web and have engendered results that far surpass current state of the art. We examine three fundamental pieces requisite for a fully functioning semantic database: dictionary encoding, RDFS inference, and query processing. We show scaling up to 512 processors (the largest configuration we had available), and the ability to process 20 billion triples completely in-memory.

99 GENERAL AND MISCELLANEOUS↗

Runtime Techniques to Enable a Highly-Scalable Global Address Space Model for Petascale Computing

Over the past decade, the trajectory to the petascale has been built on increased complexity and scale of the underlying parallel architectures. Meanwhile, software developers have struggled to provide tools that maintain the productivity of computational science teams using these new systems. In this regard, Global Address Space (GAS) programming models provide a straightforward and easy to use addressing model, which can lead to improved productivity. However, the scalability of GAS depends directly on the design and implementation of the runtime system on the target petascale distributed-memory architecture. In this paper, we describe the design, implementation, and optimization of the Aggregate Remote Memory Copy Interface (ARMCI) runtime library on the Cray XT5 2.3 PetaFLOPs computer at Oak Ridge National Laboratory. We optimized our implementation with the flow intimation technique that we have introduced in this paper. Our optimized ARMCI implementation improves scalability of both the Global Arrays programming model and a real-world chemistry application—NWChem—from small jobs up through 180,000 cores.

Computer Science↗

High Order Anchoring and Reinitialization of Level Set Function for Simulating Interface Motion

A second order interface anchoring method has been developed and used with fast sweeping algorithm for reinitialization of a level set function. The algebraic anchoring formulation ensures that the location of the actual interface is preserved, leading to better mass conservation property. It also provides high order accurate algebraic constraint for solving the Eikonal equation on a finite difference grid. Geometric properties of the interface such as surface normal and curvature are subsequently computed from the reinitialized distance function. Various analytical functions for modeling distortion in level set field are considered and accuracy of reinitialization is evaluated using first and second order anchoring schemes. It is also shown that accurate computation of interface curvature requires a high order anchor in addition to a high order fast sweeping method. Mass conservation property of reinitialization is also analyzed by considering test problems from literature including the classic Rider–Kothe single vortex problem. The formulation is suitable for efficient parallelization for both distributed memory and on-node shared memory parallel systems. In conclusion, scalability and performance of the reinitialization scheme on multiple architectures are demonstrated.

97 MATHEMATICS AND COMPUTING↗

Sparse, iterative simulation methods for one-dimensional laminar flames

Sparse, iterative simulation methods for one-dimensional laminar flames are proposed. The resulting steady and unsteady flame solvers exploit approximate Jacobians to greatly reduce the computational cost associated with matrix operations. The constant, non-unity Lewis number assumption is introduced to further reduce the computational cost. The solvers are also parallelized to reduce time-to-solution on distributed memory computer systems. Computed laminar flame speeds and species profiles for a range of chemical mechanisms (from 10 to 2878 species) are compared against a well-validated commercial code and are found to be consistent within solver tolerances. The computation times of both the unsteady and steady solutions increase only linearly with the number of species, which is a significant improvement over the quadratic or cubic scaling of existing steady-state flame solvers. For the largest mechanism tested, the steady-state flame solver is two orders of magnitude faster than commonly-used codes. The use of an approximate Jacobian is shown to reduce the rate of convergence for the steady-state solver, but does not significantly affect the domain of convergence. The steady-state solver with approximate Jacobian is thus well suited for computationally efficient laminar flame speed sweeps with large kinetic mechanisms.

42 ENGINEERING↗