Results for “parallelization” — Page 46

DOE OSTI2020

Reducing Communication in Graph Neural Network Training

Graph Neural Networks (GNNs) are powerful and flexible neural networks that use the naturally sparse connectivity information of the data. GNNs represent this connectivity as sparse matrices, which have lower arithmetic intensity and thus higher communication costs compared to dense matrices, making GNNs harder to scale to high concurrencies than convolutional or fully-connected neural networks. Here, we introduce a family of parallel algorithms for training GNNs and show that they can asymptotically reduce communication compared to previous parallel GNN training methods. We implement these algorithms, which are based on 1D, 1. 5D, 2D, and 3D sparse-dense matrix multiplication, using torch.distributed on GPU-equipped clusters. Our algorithms optimize communication across the full GNN training pipeline. We train GNNs on over a hundred GPUs on multiple datasets, including a protein network with over a billion edges.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2026

A Flexible Forwarding Scheme to Improve Latency-Bound Irregular P2P Communication in MPI

We propose an algorithm to efficiently perform latency-bound communication scenarios that consist of many small messages. In these parallel scenarios, processes typically pass around a lot of small-sized messages of a few KBs of size. Performing communication operations with P2P MPI routines or collective MPI routines (including neighborhood collectives) in such scenarios may not always yield the optimal results and may not resolve the latency bottleneck. To this end, we develop a regular structure called virtual process topology (VPT) on which the messages can be communicated in a structured and controlled manner. Using parameters of this topology, one can tune the rate of aggression in tackling the latency costs. We demonstrate that our communication algorithm is preferable to MPI P2P and collective routines for latency-bound communication and it can easily be adapted only by replacing calls to MPI routines in a parallel application. We show how to adapt existing topology-aware mapping heuristics to address the volume overhead due to communicating messages on the VPT. Moreover, we propose a novel swap-based mapping heuristic to address this overhead by optimizing the maximum volume handled by a process. Experiments on synthetic communication graphs as well as real-world applications such as parallel Canonical Polyadic sparse tensor decomposition and parallel sparse matrix-dense matrix multiplication show that our approach is a powerful way of overcoming the bottlenecks posed by sparse and latency-bound irregular communication.

communication algorithm↗

DOE OSTI2020

Characterizing Output Bottlenecks of a Production Supercomputer: Analysis and Implications

This article studies the I/O write behaviors of the Titan supercomputer and its Lustre parallel file stores under production load. The results can inform the design, deployment, and configuration of file systems along with the design of I/O software in the application, operating system, and adaptive I/O libraries.We propose a statistical benchmarking methodology to measure write performance across I/O configurations, hardware settings, and system conditions. Moreover, we introduce two relative measures to quantify the write-performance behaviors of hardware components under production load. In addition to designing experiments and benchmarking on Titan, we verify the experimental results on one real application and one real application I/O kernel, XGC and HACC IO, respectively. These two are representative and widely used to address the typical I/O behaviors of applications.In summary, we find that Titan’s I/O system is variable across the machine at fine time scales. This variability has two major implications. First, stragglers lessen the benefit of coupled I/O parallelism (striping). Peak median output bandwidths are obtained with parallel writes to many independent files, with no striping or write sharing of files across clients (compute nodes). I/O parallelism is most effective when the application—or its I/O libraries—distributes the I/O load so that each target stores files for multiple clients and each client writes files on multiple targets in a balanced way with minimal contention. Second, our results suggest that the potential benefit of dynamic adaptation is limited. In particular, it is not fruitful to attempt to identify “good locations” in the machine or in the file system: component performance is driven by transient load conditions and past performance is not a useful predictor of future performance. For example, we do not observe diurnal load patterns that are predictable.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

Simulating the Impact of Dynamic Rerouting on Metropolitan-scale Traffic Systems

The rapid introduction of mobile navigation aides that use real-time road network information to suggest alternate routes to drivers is making it more difficult for researchers and government transportation agencies to understand and predict the dynamics of congested transportation systems. Computer simulation is a key capability for these organizations to analyze hypothetical scenarios; however, the complexity of transportation systems makes it challenging for them to simulate very large geographical regions, such as multi-city metropolitan areas. In this article, we describe enhancements to the Mobiliti parallel traffic simulator to model dynamic rerouting behavior with the addition of vehicle controller actors and vehicle-to-controller reroute requests. The simulator is designed to support distributed-memory parallel execution using discrete event simulation and be scalable on high-performance computing platforms. We demonstrate the potential of the simulator by analyzing the impact of varying the population penetration rate of dynamic rerouting on the San Francisco Bay Area road network. Using high-performance parallel computing, we can simulate a day in the San Francisco Bay Area with 19 million vehicle trips with 50 percent dynamic rerouting penetration over a road network with 0.5 million nodes and 1 million links in less than three minutes. We present a sensitivity study on the dynamic rerouting parameters, discuss the simulator’s parallel scalability, and analyze system-level impacts of changing the dynamic rerouting penetration. Furthermore, we examine the varying effects on different functional classes and geographical regions and present a validation of the simulation results compared to real-world data.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

High-Level Synthesis of Irregular Applications: A Case Study on Influence Maximization

The Influence Maximization problem is the problem of identifying a small cohort of actors from a broader population that, when initially activated in a diffusion process, are expected to result in a large number of activations in the population. While the problem is known to be NP-hard, several approximation algorithms have been devised by leveraging its submodular structure. While these algorithms are theoretically efficient, they are computationally very expensive in practice. This work advances the current state-of-the-art parallelization scheme for the IMM algorithm by devising the adoption of custom hardware accelerators implemented on FPGAs by leveraging High Level Synthesis from OpenCL. We study the performance of our proposed approach by exploring optimizations tailored at improving the parallel efficiency of the accelerators and highlight their effects and limitations in accelerating complex graph analytic applications. Our experimental evaluation shows that FPGA acceleration can improve the performance of the LT diffusion model up to 1.72x for the entire application and up to 2.90x for its most important kernel with respect to a CPU only parallel execution. The FPGA acceleration of the LT model shows also a 1.54x reduction in energy consumption when compared to a parallel CPU only run.

Neff, Reece W.↗

DOE OSTI2020

Distributed Berkeley Efficient Long-Read to Long-Read Aligner and Overlapper (DiBELLA) v1.0.0

We present a parallel algorithm and scalable implementation for genome analysis, specifically the problem of finding overlaps and alignments for data from "third generation" long read sequencers. While long sequences of DNA offer enormous advantages for biological analysis and insight, current long read sequencing instruments have high error rates and therefore require different approaches to analysis than their short read counterparts. Our work focuses on an efficient distributed-memory parallelization of an accurate single-node algorithm for overlapping and aligning long reads. We achieve scalability of this irregular algorithm by addressing the competing issues of increasing parallelism, minimizing communication, constraining the memory footprint, and ensuring good load balance. The resulting application, DiBELLA, is the first distributed memory overlapper and aligner specifically designed for long reads and parallel scalability.

Ellis, Marquita↗

DOE OSTI2024

Integrating ytopt and libEnsemble to autotune OpenMC

Ytopt is a Python machine-learning-based autotuning software package developed within the ECP PROTEAS-TUNE project. The ytopt software adopts an asynchronous search framework that consists of sampling a small number of input parameter configurations and progressively fitting a surrogate model over the input-output space until exhausting the user-defined maximum number of evaluations or the wall-clock time. libEnsemble is a Python toolkit for coordinating workflows of asynchronous and dynamic ensembles of calculations across massively parallel resources developed within the ECP PETSc/TAO project. libEnsemble helps users take advantage of massively parallel resources to solve design, decision, and inference problems and expands the class of problems that can benefit from increased parallelism. In this paper we present our methodology and framework to integrate ytopt and libEnsemble to take advantage of massively parallel resources to accelerate the autotuning process. Specifically, we focus on using the proposed framework to autotune the ECP ExaSMR application OpenMC, an open source Monte Carlo particle transport code. OpenMC has seven tunable parameters some of which have large ranges such as the number of particles in-flight, which is in the range of 100,000 to 8 million, with its default setting of 1 million. Setting the proper combination of these parameter values to achieve the best performance is extremely time-consuming. Therefore, we apply the proposed framework to autotune the MPI/OpenMP offload version of OpenMC based on a user-defined metric such as the figure of merit (FoM) (particles/s) or energy efficiency energy-delay product (EDP) on Crusher at Oak Ridge Leadership Computing Facility. In conclusion, the experimental results show that we achieve the improvement up to 29.49% in FoM and up to 30.44% in EDP.

Autotuning↗

DOE OSTI2025

Fermilab's controls development with virtual accelerator

Control Systems development is often the last thing considered when designing and building new equipment, e.g. a new detector or superconducting RF LINAC; however when the new equipment is installed, it is the first thing desired to be operational for testing. Due to frequent delays in building new equipment and project deadlines, control system development and testing is often curtailed. A way to alleviate this problem is to simulate the control system, though this will be challenging for complex systems.The Fermilab PIP-II (proton improvement plan - II) project is being constructed at Fermilab to deliver $800\,MeV$ protons of $>1\,MW$ beam power to replace the present LINAC for the remainder of the existing accelerator complex. The new LINAC consists of a warm front end (WFE), 23 superconducting RF cryomodules (of 5 types), and a beam transfer line (BTL) to the existing complex.The accelerator physics group has a parallel project to create a digital twin (DT) of the PIP-II accelerator. We have coupled the EPICS controls to this DT and are developing both the DT and EPICS software in parallel. This will allow us to develop the EPICS software framework, the HMIs, sequences, high level physics applications, and other services for use in a fully functional control system.This presentation will detail the work that we have performed to date and show demonstrations of controlling and monitoring the status of the accelerator, as well as future plans for this work.

Hanlet, Pierrick [Fermilab]↗

DOE OSTI2025

Fiats: Functional inference and training for surrogates

Fiats provides a platform for research on the training and deployment of neural-network surrogate models for computational science. Fiats also supports exploring, advancing, and combining functional, object-oriented, and parallel programming patterns in Fortran 2023. As such, the Fiats name has dual expansions: “Functional Inference And Training for Surrogates” or “Fortran Inference And Training for Science.” Fiats inference and training procedures are pure and therefore satisfy a language constraint imposed on procedure invocations inside Fortran’s parallel loop construct: do concurrent. Furthermore, the Fiats training procedures are built around a do concurrent parallel reduction. Several compilers can automatically parallelize do concurrent on Central Processing Units (CPUs) or Graphics Processing Units (GPUs). Fiats thus aims to achieve performance portability through standard language mechanisms.

Rouson, Damian [Lawrence Berkeley National Laborat↗

DOE OSTI2020

Xyce Case Study

With the elimination of underground nuclear testing and declining defense budgets, science-based stockpile stewardship requires increased reliance on high performance modeling and simulation of weapon systems. Today's weapon systems are comprised of various electrical components and systems. As a result, there is a need for tools that will allow the use of massively parallel modeling and simulation techniques on high performance computers in existing and future weapons' electrical systems models. The Xyce Parallel Electronic Simulator is a SPICE (Simulation Program with Integrated Circuit Emphasis)- compatible circuit simulator designed to run on large-scale parallel computing platforms, though it can also execute efficiently on a variety of architectures including single processor workstations. As a mature platform for large-scale parallel circuit simulation, Xyce supports standard capabilities available in commercial simulators, in addition to various devices and models specific to Sandia's needs. Specifically, Xyce aids in the design and verification of electrical and electronic circuits and systems prior to weapons' manufacturing and deployment.

45 MILITARY TECHNOLOGY, WEAPONRY, AND NATIONAL DEF↗

DOE OSTI2020

Xyce Case Study

With the elimination of underground nuclear testing and declining defense budgets, science-based stockpile stewardship requires increased reliance on high performance modeling and simulation of weapon systems. Today's weapon systems are comprised of various electrical components and systems. As a result, there is a need for tools that will allow the use of massively parallel modeling and simulation techniques on high performance computers in existing and future weapons' electrical systems models. The Xyce Parallel Electronic Simulator is a SPICE (Simulation Program with Integrated Circuit Emphasis)- compatible circuit simulator designed to run on large-scale parallel computing platforms, though it can also execute efficiently on a variety of architectures including single processor workstations. As a mature platform for large-scale parallel circuit simulation, Xyce supports standard capabilities available in commercial simulators, in addition to various devices and models specific to Sandia's needs. Specifically, Xyce aids in the design and verification of electrical and electronic circuits and systems prior to weapons' manufacturing and deployment.

45 MILITARY TECHNOLOGY, WEAPONRY, AND NATIONAL DEF↗

DOE OSTI2020

PETSc Users Manual (Rev. 3.13)

This manual describes the use of PETSc for the numerical solution of partial differential equations and related problems on high-performance computers. The Portable, Extensible Toolkit for Scientific Computation (PETSc) is a suite of data structures and routines that provide the building blocks for the implementation of large-scale application codes on parallel (and serial) computers. PETSc uses the MPI standard for all message-passing communication. PETSc includes an expanding suite of parallel linear solvers, nonlinear solvers, and time integrators that may be used in application codes written in Fortran, C, C++, and Python. PETSc provides many of the mechanisms needed within parallel application codes, such as parallel matrix and vector assembly routines. The library is organized hierarchically, enabling users to employ the level of abstraction that is most appropriate for a particular problem. By using techniques of object-oriented programming, PETSc provides enormous flexibility for users. PETSc is a sophisticated set of software tools; as such, for some users it initially has a much steeper learning curve than a simple subroutine library. In particular, for individuals without some computer science background, experience programming in C, C++, python, or Fortran and experience using a debugger such as gdb or dbx, it may require a significant amount of time to take full advantage of the features that enable efficient software use. However, the power of the PETSc design and the algorithms it incorporates may make the efficient implementation of many application codes simpler than \rolling them" yourself.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

VTK-m User's' Guide (V.1.7)

High-performance computing relies on ever finer threading. Advances in processor technology include ever greater numbers of cores, hyperthreading, accelerators with integrated blocks of cores, and special vectorized instructions, all of which require more software parallelism to achieve peak performance. Traditional visualization solutions cannot support this extreme level of concurrency. Extreme scale systems require a new programming model and a fundamental change in how we design algorithms. To address these issues we created VTK-m: the visualization toolkit for multi-/many-core architectures. VTK-m supports a number of algorithms and the ability to design further algorithms through a top-down design with an emphasis on extreme parallelism. VTK-m also provides support for finding and building links across topologies, making it possible to perform operations that determine manifold surfaces, interpolate generated values, and find adjacencies. Although VTK-m provides a simplified high-level interface for programming, its template-based code removes the overhead of abstraction. VTK-m simplifies the development of parallel scientific visualization algorithms by providing a framework of supporting functionality that allows developers to focus on visualization operations. Consider the listings in Figure 1.1 that compares the size of the implementation for the Marching Cubes algorithm in VTK-m with the equivalent reference implementation in the CUDA software development kit. Because VTK-m internally manages the parallel distribution of work and data, the VTK-m implementation is shorter and easier to maintain. Additionally, VTK-m provides data abstractions not provided by other libraries that make code written in VTK-m more versatile.This book includes contributions from the VTK-m community including the VTK-m development team and the user community.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2022

The VTK-m Users' Guide (V.1.9)

High-performance computing relies on ever finer threading. Advances in processor technology include ever greater numbers of cores, hyperthreading, accelerators with integrated blocks of cores, and special vectorized instructions, all of which require more software parallelism to achieve peak performance. Traditional visualization solutions cannot support this extreme level of concurrency. Extreme scale systems require a new programming model and a fundamental change in how we design algorithms. To address these issues we created VTK-m: the visualization toolkit for multi-/many-core architectures. VTK-m supports a number of algorithms and the ability to design further algorithms through a top-down design with an emphasis on extreme parallelism. VTK-m also provides support for finding and building links across topologies, making it possible to perform operations that determine manifold surfaces, interpolate generated values, and find adjacencies. Although VTK-m provides a simplified high-level interface for programming, its template-based code removes the overhead of abstraction. VTK-m simplifies the development of parallel scientific visualization algorithms by providing a framework of supporting functionality that allows developers to focus on visualization operations. Consider the listings in Figure 1.1 that compares the size of the implementation for the Marching Cubes algorithm in VTK-m with the equivalent reference implementation in the CUDA software development kit. Because VTK-m internally manages the parallel distribution of work and data, the VTK-m implementation is shorter and easier to maintain. Additionally, VTK-m provides data abstractions not provided by other libraries that make code written in VTK-m more versatile.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2022

ExaSGD: 2022 Kernel Thrust Activities

The Kernel Thrust milestone ADSE22-407 covers the development of device-capable optimization algorithms and solvers technologies required by the ExaSGD project’s software stack in order to solve security-constrained alternating current optimal power flow (SC-ACOPF) problems on emerging exascale architectures. To this extent, in FY22 the main objective of the Kernel Thrust was (i) provide sparse optimization solver that runs efficiently on hardware accelerator devices (i.e., NVIDIA and AMD GPUs) to perform intra-node computations, (ii) strengthen the reliability and increase the performance of the mixed-dense sparse (MDS) solver of HiOp for deployment on the FY22 target architectures, Summit and Crusher, and (iii) increase performance by improving the mathematical algorithm and refining the parallel MPI-based implementation of the coarse-grain parallel solver HiOp-PriDec for capabilities deployment on the FY22 target architectures, Summit and Crusher. This document presents the developments and contributions done by the Kernels Thrust Team in FY22 toward completion of the above-mentioned objectives. These contributions progressed along four main development (sub)thrusts: (1) Design and implementation of a sparse optimization solver for use on hardware accelerators; (2) Improvement of the mathematical algorithm and of the parallel implementation of HiOp-PriDec to ensure readiness and efficient coarse-grain parallelism for FY23 target exascale machine; and (3) Support Software and Application Development Thrusts of the exaSGD project in their deployment of the project’s software stack on AMD- and NVIDIA-based architectures. The development of the sparse optimization solver (thrust 1 above) was new in FY22 and resulted in a new sparse solver in HiOp (available as of version 0.6). The second development thrust was a continuation of the efforts from FY21 and improved the mathematical algorithm and the communication strategy of the HiOp-PriDec solver. The last developement thrust is a large collaborative effort. Namely, the project’s teams from multiple labs (LLNL, PNNL, ORNL, and NREL) performed large-scale demonstration of the ExaSGD software stack, namely the optimization solvers of HiOp interfaced with the modeling front-end ExaGO and the stochastic sampler PowerScenarios. These demonstration efforts solved large-scale instances of the SC-ACOPF challenge problem of medium network sizes (10, 000-bus system) and large number of contingencies on Summit (NVIDIA accelerators) and Crusher (AMD accelerators) systems at ORNL.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

The VTK-m Users' Guide (V.2.0)

High-performance computing relies on ever finer threading. Advances in processor technology include ever greater numbers of cores, hyperthreading, accelerators with integrated blocks of cores, and special vectorized instructions, all of which require more software parallelism to achieve peak performance. Traditional visualization solutions cannot support this extreme level of concurrency. Extreme scale systems require a new programming model and a fundamental change in how we design algorithms. To address these issues we created VTK-m: the visualization toolkit for multi-/many-core architectures. VTK-m supports a number of algorithms and the ability to design further algorithms through a top-down design with an emphasis on extreme parallelism. VTK-m also provides support for finding and building links across topologies, making it possible to perform operations that determine manifold surfaces, interpolate generated values, and find adjacencies. Although VTK-m provides a simplified high-level interface for programming, its template-based code removes the overhead of abstraction. VTK-m simplifies the development of parallel scientific visualization algorithms by providing a framework of supporting functionality that allows developers to focus on visualization operations. Consider the listings in Figure 1.1 that compares the size of the implementation for the Marching Cubes algorithm in VTK-m with the equivalent reference implementation in the CUDA software development kit. Because VTK-m internally manages the parallel distribution of work and data, the VTK-m implementation is shorter and easier to maintain. Additionally, VTK-m provides data abstractions not provided by other libraries that make code written in VTK-m more versatile.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

Optimizing Distributed Training on Frontier for Large Language Models

Large language models (LLMs) have demonstrated remarkable success as foundational models, benefiting various downstream applications through fine-tuning. Loss scaling studies have demonstrated the superior performance of larger LLMs compared to their smaller counterparts. Nevertheless, training LLMs with billions of parameters poses significant challenges and requires considerable computational resources. For example, training a one trillion parameter GPT-style model on 20 trillion tokens requires a staggering 120 million exaflops. This research explores efficient distributed training strategies to extract this computation from Frontier, the world's first exascale supercomputer. We enable and investigate various model and data parallel training techniques, such as tensor parallelism, pipeline parallelism, and sharded data parallelism, to facilitate training a trillion-parameter model on Frontier. We empirically assess these techniques and their associated parameters to determine their impact on memory footprint, communication latency, and GPU's computational efficiency. We analyze the complex interplay among these techniques and find a strategy to combine them to achieve high throughput through hyperparameter tuning. We have identified efficient strategies for training large LLMs of varying sizes through empirical analysis and hyperparameter tuning. For 22 Billion, 175 Billion, and 1 Trillion parameters, we achieved GPU throughputs of 38.38%, 36.14%, and 31.96%, respectively. For the training of the 175 Billion parameter model and the 1 Trillion parameter model, we achieved 100% weak scaling efficiency on 1024 and 3072 Mi250X GPUs, respectively. We also achieved strong scaling efficiencies of 89% and 87% for these two models. We trained these models only tens of iterations instead of training till completion.

Yin, Junqi↗

DOE OSTI2022

Caffeine: CoArray Fortran Framework of Efficient Interfaces to Network Environments

This paper provides an introduction to the CoArray Fortran Framework of Efficient Interfaces to Network Environments (Caffeine), a parallel runtime library built atop the GASNet-EX exascale networking library. Caffeine leverages several non-parallel Fortran features to write type- and rank-agnostic interfaces and corresponding procedure definitions that support parallel Fortran 2018 features, including communication, collective operations, and related services. One major goal is to develop a runtime library that can eventually be considered for adoption by LLVM Flang, enabling that compiler to support the parallel features of Fortran. The paper describes the motivations behind Caffeine's design and implementation decisions, details the current state of Caffeine's development, and previews future work. We explain how the design and implementation offer benefits related to software sustainability by lowering the barrier to user contributions, reducing complexity through the use of Fortran 2018 C-interoperability features, and high performance through the use of a lightweight communication substrate.

Rouson, Damian↗