Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Data parallelism”

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 109 records · Page 6
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

The Case for Strong Scaling in Deep Learning: Training Large 3D CNNs with Hybrid Parallelism

Here, we present scalable hybrid-parallel algorithms for training large-scale 3D convolutional neural networks. Deep learning-based emerging scientific workflows often require model training with large, high-dimensional samples, which can make training much more costly and even infeasible due to excessive memory usage. We solve these challenges by extensively applying hybrid parallelism throughout the end-to-end training pipeline, including both computations and I/O. Our hybrid-parallel algorithm extends the standard data parallelism with spatial parallelism, which partitions a single sample in the spatial domain, realizing strong scaling beyond the mini-batch dimension with a larger aggregated memory capacity. We evaluate our proposed training algorithms with two challenging 3D CNNs, CosmoFlow and 3D U-Net. Our comprehensive performance studies show that good weak and strong scaling can be achieved for both networks using up to 2K GPUs. More importantly, we enable training of CosmoFlow with much larger samples than previously possible, realizing an order-of-magnitude improvement in prediction accuracy.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

HydraGNN v3.0

New or improved capabilities included in v3.0 release are as follows: 1. Enhancement in message passing layers through generalization of the class inheritance to enable the inclusion of a broader set of message passing policies Inclusion of equivariant message passing layers from the original implementations of: SchNet (https://pubs.aip.org/aip/jcp/article/148/24/241722/962591/SchNet-A-deep-learning-architecture-for-molecules); DimeNet++ (https://arxiv.org/abs/2011.14115); EGNN models (https://arxiv.org/pdf/2102.09844.pdf) 2. Restructuring of class inheritance for data management 3. Support of DDStore https://github.com/ORNL/DDStore capabilities for improved distributed data parallelism on large volumes of data that cannot fit on intra-node memory capacities 4. Large-scale system support for OLCF-Crusher and OLCF-Frontier

Lupo Pasini, Massimiliano [Oak Ridge National Labo↗
NASA NTRS1994

An overview of the Opus language and runtime system

We have recently introduced a new language, called Opus, which provides a set of Fortran language extensions that allow for integrated support of task and data parallelism. lt also provides shared data abstractions (SDA's) as a method for communication and synchronization among these tasks. In this paper, we first provide a brief description of the language features and then focus on both the language-dependent and language-independent parts of the runtime system that support the language. The language-independent portion of the runtime system supports lightweight threads across multiple address spaces, and is built upon existing lightweight thread and communication systems. The language-dependent portion of the runtime system supports conditional invocation of SDA methods and distributed SDA argument handling.

Mehrotra, Piyush↗
NASA NTRS1997

Solution of Aeroacoustic Problems by a Nonlinear, Hybrid Method

Category 1, problem 3 (scattering of sound by a sphere) and category 2, problem 1 (spherical source in a cylindrical duct subject to uniform flow) are solved in generalized coordinates using the nonlinear Euler equations together with nonreflecting boundary conditions. A temporally and spatially fourth-order accurate finite-difference, Runge-Kutta time-marching technique is employed for the near-field calculations and a Kirchhoff method is employed for the prediction of far-field sound. Computations are all performed on parallel processors using the data-parallel paradigm.

Oezyoeruek, Yusuf↗
DOE OSTI2020

Scalable Data-Intensive Geocomputation: A Design for Real-Time Continental Flood Inundation Mapping

The convergence of data-intensive and extreme-scale computing enables an integrated software and data ecosystem for scientific discovery. Developments in this realm will fuel transformative research in data-driven interdisciplinary domains. Geocomputation provides computing paradigms in Geographic Information Systems (GIS) for interactive computing of geographic data, processes, models, and maps. Because GIS is data-driven, the computational scalability of a geocomputation workflow is directly related to the scale of the GIS data layers, their resolution and extent, as well as the velocity of the geo-located data streams to be processed. Unique in high user interactivity and low end-to-end latency requirements, geocomputation applications will dramatically benefit from the convergence of high-end data analytics (HDA) and high-performance computing (HPC). The application level challenge, however, is to identify and eliminate computational bottlenecks that arise along a geocomputation workflow. Indeed, poor scalability at any of the workflow components is detrimental to the entire end-to-end pipeline. Here, we study a large geocomputation use case in flood inundation mapping that handles multiple national-scale geospatial datasets and targets low end-to-end latency. We discuss benefits and challenges for harnessing both HDA and HPC for data-intensive geospatial data processing and intensive numerical modeling of geographic processes. We propose an HDA+HPC geocomputation architecture design that couples HDA (e.g., Spark)-based spatial data handling and HPC-based parallel data modeling. Key techniques for coupling HDA and HPC to bridge the two different software stacks are reviewed and discussed.

Liu, Yan↗
NASA NTRS1997

Geopotential Error Analysis from Satellite Gradiometer and Global Positioning System Observables on Parallel Architecture

The recovery of a high resolution geopotential from satellite gradiometer observations motivates the examination of high performance computational techniques. The primary subject matter addresses specifically the use of satellite gradiometer and GPS observations to form and invert the normal matrix associated with a large degree and order geopotential solution. Memory resident and out-of-core parallel linear algebra techniques along with data parallel batch algorithms form the foundation of the least squares application structure. A secondary topic includes the adoption of object oriented programming techniques to enhance modularity and reusability of code. Applications implementing the parallel and object oriented methods successfully calculate the degree variance for a degree and order 110 geopotential solution on 32 processors of the Cray T3E. The memory resident gradiometer application exhibits an overall application performance of 5.4 Gflops, and the out-of-core linear solver exhibits an overall performance of 2.4 Gflops. The combination solution derived from a sun synchronous gradiometer orbit produce average geoid height variances of 17 millimeters.

Schutz, Bob E.↗
NASA NTRS2012

Gigabit Ethernet Asynchronous Clock Compensation FIFO

Clock compensation for Gigabit Ethernet is necessary because the clock recovered from the 1.25 Gb/s serial data stream has the potential to be 200 ppm slower or faster than the system clock. The serial data is converted to 10-bit parallel data at a 125 MHz rate on a clock recovered from the serial data stream. This recovered data needs to be processed by a system clock that is also running at a nominal rate of 125 MHz, but not synchronous to the recovered clock. To cross clock domains, an asynchronous FIFO (first-in-first-out) is used, with the write pointer (wprt) in the recovered clock domain and the read pointer (rptr) in the system clock domain. Because the clocks are generated from separate sources, there is potential for FIFO overflow or underflow. Clock compensation in Gigabit Ethernet is possible by taking advantage of the protocol data stream features. There are two distinct data streams that occur in Gigabit Ethernet where identical data is transmitted for a period of time. The first is configuration, which happens during auto-negotiation. The second is idle, which occurs at the end of auto-negotiation and between every packet. The identical data in the FIFO can be repeated by decrementing the read pointer, thus compensating for a FIFO that is draining too fast. The identical data in the FIFO can also be skipped by incrementing the read pointer, which compensates for a FIFO draining too slowly. The unique and novel features of this FIFO are that it works in both the idle stream and the configuration streams. The increment or decrement of the read pointer is different in the idle and compensation streams to preserve disparity. Another unique feature is that the read pointer to write pointer difference range changes between compensation and idle to minimize FIFO latency during packet transmission.

Duhachek, Jeff↗
DOE OSTI2020

Improving Performance of M-to-N Processing and Data Redistribution in In Transit Analysis and Visualization

In an in transit setting, a parallel data producer, such as a numerical simulation, runs on one set of ranks M, while a data consumer, such as a parallel visualization application, runs on a different set of ranks N. One of the central challenges in this in transit setting is to determine the mapping of data from the set of M producer ranks to the set of N consumer ranks. This is a challenging problem for several reasons, such as the producer and consumer codes potentially having different scaling characteristics and different data models. The resulting mapping from M to N ranks can have a significant impact on aggregate application performance. In this work, we present an approach for performing this M-to-N mapping in a way that has broad applicability across a diversity of data producer and consumer applications. We evaluate its design and performance with a study that runs at high concurrency on a modern HPC platform. By leveraging design characteristics, which facilitate an “intelligent” mapping from M-to-N, we observe significant performance gains are possible in terms of several different metrics, including time-to-solution and amount of data moved.

Loring, Burlen↗
DOE OSTI2020

Improving Performance of M-to-N Processing and Data Redistribution in In Transit Analysis and Visualization

In an in transit setting, a parallel data producer, such as a numerical simulation, runs on one set of ranks M, while a data consumer, such as a parallel visualization application, runs on a different set of ranks N: One of the central challenges in this in transit setting is to determine the mapping of data from the set of M producer ranks to the set of N consumer ranks. This is a challenging problem for several reasons, such as the producer and consumer codes potentially having different scaling characteristics and different data models. The resulting mapping from M to N ranks can have a significant impact on aggregate application performance. In this work, we present an approach for performing this M-to-N mapping in a way that has broad applicability across a diversity of data producer and consumer applications. We evaluate its design and performance with a study that runs at high concurrency on a modern HPC platform. By leveraging design characteristics, which facilitate an ''intelligent'' mapping from M-to-N, we observe significant performance gains are possible in terms of several different metrics, including time-to-solution and amount of data moved.

Loring, Burlen↗
DOE OSTI2021

Minimizing development costs for efficient many-core visualization using MCD3

Scientific visualization software increasingly needs to support many-core architectures. However, development time is a significant challenge due to the breadth and diversity of both visualization algorithms and architectures. With this work, we introduce a development environment for visualization algorithms on many-core devices that extends the traditional data-parallel primitive (DPP) approach with several existing constructs and an important new construct: meta-DPPs. We refer to our approach as MCD 3 — Meta-DPPs, Convenience routines, Data management, DPPs, and Devices. The twin goals of MCD 3 are to reduce developer time and to deliver efficient performance on many-core architectures, and our evaluation considers both of these goals. For development time, we study 57 algorithms implemented in the VTK-m software library and determine that MCD 3 leads to significant savings. For efficient performance, we survey ten studies looking at individual algorithms and determine that the MCD 3 hardware-agnostic approach leads to performance comparable to hardware-specific approaches: sometimes better, sometimes worse, and better in the aggregate. In total, we find that MCD 3 is an effective approach for scientific visualization libraries to support many-core architectures.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1989

Fast Asynchronous Data Communication Via Fiber Optics

Transmitter and receiver devised for asynchronous digital communication via optical fiber at rates above 100 Mb/s. Transmitter converts parallel data to serial for high-speed transmission; receiver recovers clock signal and converts data back to parallel. No phase-lock loops used. New receiver design avoids over-sampling altogether. Local sampling oscillator operating nominally at clock frequency generates N clock signals of equally spaced phase, used to clock incoming data into N separate shift registers.

Bergman, Larry A.↗
NASA NTRS1995

A portable MPI-based parallel vector template library

This paper discusses the design and implementation of a polymorphic collection library for distributed address-space parallel computers. The library provides a data-parallel programming model for C++ by providing three main components: a single generic collection class, generic algorithms over collections, and generic algebraic combining functions. Collection elements are the fourth component of a program written using the library and may be either of the built-in types of C or of user-defined types. Many ideas are borrowed from the Standard Template Library (STL) of C++, although a restricted programming model is proposed because of the distributed address-space memory model assumed. Whereas the STL provides standard collections and implementations of algorithms for uniprocessors, this paper advocates standardizing interfaces that may be customized for different parallel computers. Just as the STL attempts to increase programmer productivity through code reuse, a similar standard for parallel computers could provide programmers with a standard set of algorithms portable across many different architectures. The efficacy of this approach is verified by examining performance data collected from an initial implementation of the library running on an IBM SP-2 and an Intel Paragon.

Sheffler, Thomas J.↗
NASA NTRS1995

A Portable MPI-Based Parallel Vector Template Library

This paper discusses the design and implementation of a polymorphic collection library for distributed address-space parallel computers. The library provides a data-parallel programming model for C + + by providing three main components: a single generic collection class, generic algorithms over collections, and generic algebraic combining functions. Collection elements are the fourth component of a program written using the library and may be either of the built-in types of c or of user-defined types. Many ideas are borrowed from the Standard Template Library (STL) of C++, although a restricted programming model is proposed because of the distributed address-space memory model assumed. Whereas the STL provides standard collections and implementations of algorithms for uniprocessors, this paper advocates standardizing interfaces that may be customized for different parallel computers. Just as the STL attempts to increase programmer productivity through code reuse, a similar standard for parallel computers could provide programmers with a standard set of algorithms portable across many different architectures. The efficacy of this approach is verified by examining performance data collected from an initial implementation of the library running on an IBM SP-2 and an Intel Paragon.

Sheffler, Thomas J.↗
NASA NTRS1981

28-Channel rotary transformer

Transformer transmits power and digital data across rotating interface. Array has many parallel data channels, each with potential l megabaud data rate. Ferrite-cored transformers are spaced along rotor; airgap between them reduces crosstalk.

Mclyman, W. T.↗
DOE OSTI2024

Efficient distributed continual learning for steering experiments in real-time

Deep learning has emerged as a powerful method for extracting valuable information from large volumes of data. However, when new training data arrives continuously (i.e., is not fully available from the beginning), incremental training suffers from catastrophic forgetting (i.e., new patterns are reinforced at the expense of previously acquired knowledge). Training from scratch each time new training data becomes available would result in extremely long training times and massive data accumulation. Rehearsal-based continual learning has shown promise for addressing the catastrophic forgetting challenge, but research to date has not addressed performance and scalability. To fill this gap, we propose an approach based on a distributed rehearsal buffer that efficiently complements data-parallel training on multiple GPUs to achieve high accuracy, short runtime, and scalability. It leverages a set of buffers (local to each GPU) and uses several asynchronous techniques for updating these local buffers in an embarrassingly parallel fashion, all while handling the communication overheads necessary to augment input minibatches using unbiased, global sampling. We further propose a generalization of rehearsal buffers to support both classification and generative learning tasks, as well as more advanced rehearsal strategies (notably Dark Experience Replay, leveraging knowledge distillation). We illustrate this approach with a real-life HPC streaming application from the domain of ptychographic image reconstruction. Furthermore, we run extensive experiments on up to 128 GPUs of the ThetaGPU supercomputer to compare our approach with baselines representative of training-from-scratch (the upper bound in terms of accuracy) and incremental training (the lower bound). Results show that rehearsal-based continual learning achieves a top-5 validation accuracy close to the upper bound, while simultaneously exhibiting a runtime close to the lower bound.

Asynchronous data management↗
DOE OSTI2020

Calibration and modification of the constitutive model for the mechanical response of P91

The objective of this first milestone was to calibrate and if necessary, modify a constitutive model predicting the primary and secondary creep response of P91. The overall path to be followed consists of first calibrating this constitutive model against experimental data available from the literature (this milestone). In parallel, data acquired by indentation tests will be used to calibrate the model (this involves transforming the indentation data into equivalent uniaxial tests). An overall comparison of model parameters will allow assessing to what extent these rapid tests can be used to calibrate models predicting the overall mechanical response of the bulk. Given that COVID naturally hindered the gathering of new experimental data, the choice was made to ensure that LANL would have sufficient funds in year two to perform the bulk of the new calibration work. The following report details (i) how and which of the bulk experimental data was selected, (ii) constitutive model formulation, (iii) predictions of the constitutive model: effects of imposed stress, temperature and precipitate content, (iv) future work.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
NASA NTRS1995

Ropes: Support for collective opertions among distributed threads

Lightweight threads are becoming increasingly useful in supporting parallelism and asynchronous control structures in applications and language implementations. Recently, systems have been designed and implemented to support interprocessor communication between lightweight threads so that threads can be exploited in a distributed memory system. Their use, in this setting, has been largely restricted to supporting latency hiding techniques and functional parallelism within a single application. However, to execute data parallel codes independent of other threads in the system, collective operations and relative indexing among threads are required. This paper describes the design of ropes: a scoping mechanism for collective operations and relative indexing among threads. We present the design of ropes in the context of the Chant system, and provide performance results evaluating our initial design decisions.

Haines, Matthew↗
NASA NTRS2004

Means, Variability and Trends of Precipitation in the Global Climate as Determined by the 25-year GEWEWGPCP Data Set

The Global Precipitation Climatology Project (GPCP) 25-year precipitation data set is used as a basis to evaluate the mean state, variability and trends (or inter-decadal changes) of global and regional scales of precipitation. The uncertainties of these characteristics of the data set are evaluated by examination of other, parallel data sets and examination of shorter periods with higher quality data (e.g., TRMM). The global and regional means are assessed for uncertainty by comparing with other satellite and gauge data sets, both globally and regionally. The GPCP global mean of 2.6 mdday is divided into values of ocean and land and major latitude bands (Tropics, mid-latitudes, etc.). Seasonal variations globally and by region are shown and uncertainties estimated. The variability of precipitation year-to-year is shown to be related to ENS0 variations and volcanoes and is evaluated in relation to the overall lack of a significant global trend. The GPCP data set necessarily has a heterogeneous time series of input data sources, so part of the assessment described above is to test the initial results for potential influence by major data boundaries in the record.

Adler, R. F.↗