Search NASA⌕ Search

SEARCH · Search NASA

Results for “parallel computer architecture”

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 613 records · Page 34

A Generalized Framework for Auto-tuning Stencil Computations

This work introduces a generalized framework for automatically tuning stencil computations to achieve superior performance on a broad range of multicore architectures. Stencil (nearest-neighbor) based kernels constitute the core of many important scientific applications involving block-structured grids. Auto-tuning systems search over optimization strategies to find the combination of tunable parameters that maximizes computational efficiency for a given algorithmic kernel. Although the auto-tuning strategy has been successfully applied to libraries, generalized stencil kernels are not amenable to packaging as libraries. Studied kernels in this work include both memory-bound kernels as well as a computation-bound bilateral filtering kernel. We introduce a generalized stencil auto-tuning framework that takes a straightforward Fortran expression of a stencil kernel and automatically generates tuned implementations of the kernel in C or Fortran to achieve performance portability across diverse computer architectures.

97 MATHEMATICS AND COMPUTING↗

Roadmap on electronic structure codes in the exascale era

Electronic structure calculations have been instrumental in providing many important insights into a range of physical and chemical properties of various molecular and solid-state systems. Their importance to various fields, including materials science, chemical sciences, computational chemistry and device physics, is underscored by the large fraction of available public supercomputing resources devoted to these calculations. As we enter the exascale era, exciting new opportunities to increase simulation numbers, sizes, and accuracies present themselves. In order to realize these promises, the community of electronic structure software developers will however first have to tackle a number of challenges pertaining to the efficient use of new architectures that will rely heavily on massive parallelism and hardware accelerators. This roadmap provides a broad overview of the state-of-the-art in electronic structure calculations and of the various new directions being pursued by the community. It covers 14 electronic structure codes, presenting their current status, their development priorities over the next five years, and their plans towards tackling the challenges and leveraging the opportunities presented by the advent of exascale computing.

36 MATERIALS SCIENCE↗

Massively parallel and universal approximation of nonlinear functions using diffractive processors

Nonlinear computation is essential for a wide range of information processing tasks, yet implementing nonlinear functions using optical systems remains a challenge due to the weak and power-intensive nature of optical nonlinearities. Overcoming this limitation without relying on nonlinear optical materials could unlock unprecedented opportunities for ultrafast and parallel optical computing systems. Here, we demonstrate that large-scale nonlinear computation can be performed using linear optics through optimized diffractive processors composed of passive phase-only surfaces. In this framework, the input variables of nonlinear functions are encoded into the phase of an optical wavefront—e.g., via a spatial light modulator (SLM)—and transformed by an optimized diffractive structure with spatially varying point-spread functions to yield output intensities that approximate a large set of unique nonlinear functions–all in parallel. We provide proof establishing that this architecture serves as a universal function approximator for an arbitrary set of bandlimited nonlinear functions, also covering wavelength-multiplexed nonlinear functions as well as multi-variate and complex-valued functions that are all-optically cascadable. Our analysis also indicates the successful approximation of typical nonlinear activation functions commonly used in neural networks, including the sigmoid, tanh, ReLU (rectified linear unit), and softplus. We numerically demonstrate the parallel computation of one million distinct nonlinear functions, accurately executed at wavelength-scale spatial density at the output of a diffractive optical processor. Furthermore, we experimentally validated this framework using in situ optical learning and approximated 35 unique nonlinear functions in a single shot using a compact setup consisting of an SLM and an image sensor. These results establish diffractive optical processors as a scalable platform for massively parallel universal nonlinear function approximation, paving the way for new capabilities in analog optical computing based on linear materials.

Rahman, Md Sadman Sakib [University of California,↗

Stochastic simulation of electron avalanches on supercomputers

In the paper, we present a three-dimensional parallel Monte Carlo algorithm named ELSHOW which is developed for simulation of electron avalanches in gases. Parallel implementation of the ELSHOW was made on supercomputers with different architectures (massive parallel and hybrid ones). Using the ELSHOW, calculations of such integral characteristics as the number of particles in an avalanche, the coefficient of impact ionization, the drift velocity, and the others were made. Also, special precise computations were made to select an appropriate size of the time step using the technique of dependent statistical tests. Particularly, the algorithm consists of special methods of distribution modeling, a lexicographic implementation scheme for “branching” of trajectories, justified estimation of functionals. A comparison of the obtained results for nitrogen with previously published theoretical and experimental data was made.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Hardware Architectures for Data-Intensive Computing Problems: A Case Study for String Matching

DNA analysis is an emerging application of high performance bioinformatic. Modern sequencing machinery are able to provide, in few hours, large input streams of data, which needs to be matched against exponentially growing databases of known fragments. The ability to recognize these patterns effectively and fastly may allow extending the scale and the reach of the investigations performed by biology scientists. Aho-Corasick is an exact, multiple pattern matching algorithm often at the base of this application. High performance systems are a promising platform to accelerate this algorithm, which is computationally intensive but also inherently parallel. Nowadays, high performance systems also include heterogeneous processing elements, such as Graphic Processing Units (GPUs), to further accelerate parallel algorithms. Unfortunately, the Aho-Corasick algorithm exhibits large performance variability, depending on the size of the input streams, on the number of patterns to search and on the number of matches, and poses significant challenges on current high performance software and hardware implementations. An adequate mapping of the algorithm on the target architecture, coping with the limit of the underlining hardware, is required to reach the desired high throughputs. In this paper, we discuss the implementation of the Aho-Corasick algorithm for GPU-accelerated high performance systems. We present an optimized implementation of Aho-Corasick for GPUs and discuss its tradeoffs on the Tesla T10 and he new Tesla T20 (codename Fermi) GPUs. We then integrate the optimized GPU code, respectively, in a MPI-based and in a pthreads-based load balancer to enable execution of the algorithm on clusters and large sharedmemory multiprocessors (SMPs) accelerated with multiple GPUs.

Tumeo, Antonino↗

Efficient pattern matching on GPUs for intrusion detection systems

Pattern matching is at the core of many security applications, like Network Intrusion Detection Systems (NIDS), spam filters and virus scanner. The always growing traffic on networks requires the ability to recognize potentially malicious signatures effectively, fastly and possibly in real time, without afftecting the performance and the latencies of the connections. Unfortunately, pattern matching is a computationally intensive procedure which poses significant challenges on current software and hardware implementations. Graphic Processing Units (GPU) have become an interesting target for such high-througput applications, but the algorithms and the data structures need to be redesigned to be parallelized and adapted to the underlining hardware, coping with the limitations imposed by these architectures. In this paper we present an efficient implementation of the Aho-Corasick pattern matching algorithm on GPU, showing how we progressively redesigned the algorithm and the data structures to fit on the architecture and comparing it with equivalent implementations on the CPU and with previous work. We show that with realistic TCP-IP workloads and signatures, our implementation obtains a speedup of 6.5 with respect to CPU implementations and of two times when compared to previous GPU solutions.

97 MATHEMATICS AND COMPUTING↗

Studying CPU and memory utilization of applications on Fujitsu A64FX and Nvidia Grace Superchip

ARM-based manycore CPU architectures are well-positioned to provide the rising memory throughput requirements of modern data intensive scientific applications in High Performance Computing (HPC). The Fujitsu A64FX CPU platform is based on the ARM v8.2A architecture, and is the processor of the flagship Japanese supercomputer - "Fugaku", which was previously ranked as the #1 supercomputer in the world according to the Top500 list. The Nvidia Grace superchip features 144 Neoverse V2 cores based on the ARMv9 architecture with 4x128b SVE2, providing exceptional computational power. The chip supports up to 480GB of memory, making it ideal for AI, machine learning, and scientific computing workloads. In this paper, we conduct a thorough performance exploration of a variety of parallel bandwidth-sensitive benchmarks and applications compiled with the native Fujitsu compiler on a Fugaku A64FX compute node and ARM (LLVM) Compiler on an NVIDIA Grace superchip compute node, engaging all the computational cores per cluster using OpenMP multithreading (assuming the cores can drive the available bandwidth). Our ultimate goals are to study the resource utilization of scientific applications and benchmarks on A64FX and Grace superchip, considering graph application scenarios ( GAP Benchmark suite) and eleven appli- cation proxies from the Rodinia heterogeneous benchmark suite (considering domains such as Data Mining, Bioinformatics, Fluid Dynamics, Pattern Recognition, etc.). Through exhaustive performance monitoring, we quantify the resource utilization of diverse OpenMP-based HPC applications on both the Fujitsu A64FX and the Nvidia Grace Superchip platforms.

benchmarking, Performance Analysis, High performan↗

Structural Simluation Toolkit (SST) v.12.0

The Structural Simulation Toolkit (SST) was developed to explore innovations in highly concurrent computing systems where the instruction set architecture (ISA), micro-architecture, and memory interact with the programming model and communications system. The package provides a fully modular design for extensive exploration of an individual system parameter as well as a parallel simulation environment based on message passing interface (MPI) which enable a high level of performance as well as the ability to look at large systems. Sandia National Laboratories is a multimission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525.

Rodrigues, ArunF.↗

Scalable Track Detection in SAR CCD Images

Existing methods to detect vehicle tracks in coherent change detection images, a product of combining two synthetic aperture radar images ta ken at different times of the same scene, rely on simple, fast models to label track pixels. These models, however, are often too simple to capture natural track features such as continuity and parallelism. We present a simple convolutional network architecture consisting of a series of 3-by-3 convolutions to detect tracks. The network is trained end-to-end to learn natural track features entirely from data. The network is computationally efficient and improves the F-score on a standard dataset to 0.988, up fr om 0.907 obtained by the current state-of-the-art method.

47 OTHER INSTRUMENTATION↗

TORCH Computational Reference Kernels - A Testbed for Computer Science Research

For decades, computer scientists have sought guidance on how to evolve architectures, languages, and programming models in order to improve application performance, efficiency, and productivity. Unfortunately, without overarching advice about future directions in these areas, individual guidance is inferred from the existing software/hardware ecosystem, and each discipline often conducts their research independently assuming all other technologies remain fixed. In today's rapidly evolving world of on-chip parallelism, isolated and iterative improvements to performance may miss superior solutions in the same way gradient descent optimization techniques may get stuck in local minima. To combat this, we present TORCH: A Testbed for Optimization ResearCH. These computational reference kernels define the core problems of interest in scientific computing without mandating a specific language, algorithm, programming model, or implementation. To compliment the kernel (problem) definitions, we provide a set of algorithmically-expressed verification tests that can be used to verify a hardware/software co-designed solution produces an acceptable answer. Finally, to provide some illumination as to how researchers have implemented solutions to these problems in the past, we provide a set of reference implementations in C and MATLAB.

97 MATHEMATICS AND COMPUTING↗

Characterizing the performance of node-aware strategies for irregular point-to-point communication on heterogeneous architectures

Supercomputer architectures are trending toward higher computational throughput due to the inclusion of heterogeneous compute nodes. These multi-GPU nodes increase on-node computational efficiency, while also increasing the amount of data to be communicated and the number of potential data flow paths. In this work, we characterize the performance of irregular point-to-point communication with MPI on heterogeneous compute environments through performance modeling, demonstrating the limitations of standard communication strategies for both device-aware and staging-through-host communication techniques. Presented models suggest staging communicated data through host processes then using node-aware communication strategies for high inter-node message counts. Notably, the models also predict that node-aware communication utilizing all available CPU cores to communicate inter-node data leads to the most performant strategy when communicating with a high number of nodes. Furthermore, model validation is provided via a case study of irregular point-to-point communication patterns in distributed sparse matrix–vector products. Importantly, we include a discussion on the implications model predictions have on communication strategy design for emerging supercomputer architectures.

97 MATHEMATICS AND COMPUTING↗

A PARALLEL MONTE CARLO CODE FOR SIMULATING COLLISIONAL N-BODY SYSTEMS

We present a new parallel code for computing the dynamical evolution of collisional N-body systems with up to N {approx} 10{sup 7} particles. Our code is based on the Henon Monte Carlo method for solving the Fokker-Planck equation, and makes assumptions of spherical symmetry and dynamical equilibrium. The principal algorithmic developments involve optimizing data structures and the introduction of a parallel random number generation scheme as well as a parallel sorting algorithm required to find nearest neighbors for interactions and to compute the gravitational potential. The new algorithms we introduce along with our choice of decomposition scheme minimize communication costs and ensure optimal distribution of data and workload among the processing units. Our implementation uses the Message Passing Interface library for communication, which makes it portable to many different supercomputing architectures. We validate the code by calculating the evolution of clusters with initial Plummer distribution functions up to core collapse with the number of stars, N, spanning three orders of magnitude from 10{sup 5} to 10{sup 7}. We find that our results are in good agreement with self-similar core-collapse solutions, and the core-collapse times generally agree with expectations from the literature. Also, we observe good total energy conservation, within {approx}< 0.04% throughout all simulations. We analyze the performance of the code, and demonstrate near-linear scaling of the runtime with the number of processors up to 64 processors for N = 10{sup 5}, 128 for N = 10{sup 6} and 256 for N = 10{sup 7}. The runtime reaches saturation with the addition of processors beyond these limits, which is a characteristic of the parallel sorting algorithm. The resulting maximum speedups we achieve are approximately 60 Multiplication-Sign , 100 Multiplication-Sign , and 220 Multiplication-Sign , respectively.

79 ASTRONOMY AND ASTROPHYSICS↗

FLEET: Flexible Efficient Ensemble Training for Heterogeneous Deep Neural Networks

Parallel training of an ensemble of Deep Neural Networks (DNN) on a cluster of nodes is an effective approach to shorten the process of neural network architecture search and hyper-parameter tuning for a given learning task. Prior efforts have shown that data sharing, where the common preprocessing operation is shared across the DNN training pipelines, saves computational resources and improves pipeline efficiency. Data sharing strategy, however, performs poorly for a heterogeneous set of DNNs where each DNN has varying computational needs and thus different training rate and convergence speed. This paper proposes FLEET, a flexible ensemble DNN training framework for efficiently training a heterogeneous set of DNNs. We build FLEET via several technical innovations. We theoretically prove that an optimal resource allocation is NP-hard and propose a greedy algorithm to efficiently allocate resources for training each DNN with data sharing. We integrate data-parallel DNN training into ensemble training to mitigate the differences in training rates and introduce checkpointing into this context to address the issue of different convergence speeds. Experiments show that FLEET significantly improves the training efficiency of DNN ensembles without compromising the quality of the result.

Guan, Hui↗

Collaborative Research: A Model of Partially Ionized Plasma Flows with Kinetic Treatment of Neutral Atoms and Nonthermal Ions

Interactions of flows of partially ionized, magnetized plasma are frequently accompanied by the presence of both thermal and non-thermal (pickup) ion components. Such interactions cannot be modeled using traditional MHD equations and require more advanced approaches to treat them. If a nonthermal component of ions is formed due to charge exchange and collisions between the thermal (core) ions and neutrals, it experiences the action of magnetic field, its distribution function is isotropized, and it soon acquires the velocity of the ambient plasma without being thermodynamically equilibrated. This situation, e. g., takes place in the outer heliosphere –- the part of interstellar space beyond the solar system whose properties are determined by the solar wind interaction with the local interstellar medium. This is also possible in laboratory, at million degrees and above, when plasma is conducting electricity far too well, which makes Ohmic heating ineffective. To attain the target temperatures one needs additional heating eventually playing a dominant role. Among such sources is a so-called neutral particle beam heating. This is a wide-spread technique (Joint European Torus and International Thermonuclear Experimental Reactor experiments) based on the injection of powerful beams of neutral atoms into ohmically preheated plasma. In this project we have investigated the energy and density separation between the thermal and nonthermal components in the solar wind and interstellar plasmas. A new model has been developed in which we solve the ideal MHD equations for mixture of all ions and the kinetic Boltzmann equation to describe the transport of neutral atoms. As a separate capability, we can treat the flow of neutral atoms in a multi-component fashion, where neutral atoms born in each thermodynamically distinct region are governed by the Euler gas dynamic equations. We also describe the behavior of pickup ions either kinetically, using the Fokker--Planck equation, or as a separate fluid. Our numerical simulations have demonstrated that pickup ions play a major role in the interaction of the solar wind and (partially ionized) interstellar medium plasmas. Our teams have investigated the stability of the surface (the heliopause) that separates the solar wind from the local interstellar medium, the transport of galactic cosmic rays, the properties of the heliotail flow, and modifications to the bow wave in front of the heliopause due to charge exchange between the neutral H atoms born in the solar wind and interstellar ions. Modeling results have been validated against observational data, such as obtained by the Interstellar Boundary Explorer (IBEX), and made it possible to shed light on the structure of energetic neutral atom maps created by this spacecraft.. We have also demonstrated that charge-exchange modulated heliosphere is a source of anisotropy of the multi-TeV cosmic ray flux observed in a number of Earth-bound air shower experiments. Newly developed codes are implemented within a Multi-Scale Fluid-Kinetic Simulation Suite (MS-FLUKSS), a publicly available code being developed by our team for over 12 years. MS-FLUKSS scales well up to 160,000 computing cores and has been ported on major supercomputers in the country. Efficient parallelization and data choreography in the continuum simulation modules are provided by Chombo, an adaptive mesh refinement framework managed by Phillip Colella's team at LBNL. We have implemented in-house, hybrid (MPI+OpenMP) parallelization of the kinetic modules that solve the Boltzmann equation with a Monte Carlo method. Currently, the kinetic modules are being rewritten to take advantage of the modern CPU-GPU supercomputer architecture. The scope of the project allowed us to enhance plasma research and education in such broad, multidisciplinary field as physics of partially ionized plasma and its application to space physics and fusion science. Besides the impact on the modeling of complex physical systems, our approach to computational resource management for complex codes utilizing multiple algorithm technologies appears to be a major advance on current approaches. The development of sophisticated resource management will be essential for all future modeling efforts that incorporate a diversity of scales and physical processes. Our effort provided leadership in promoting computational science and plasma physics within the UAH and FIT campuses and, through the training of a broad spectrum of scientists and engineers, fostering new technologies across the country.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

A Review of Lightweight Thread Approaches for High Performance Computing

High-level, directive-based solutions are becoming the programming models (PMs) of the multi/many-core architectures. Several solutions relying on operating system (OS) threads perfectly work with a moderate number of cores. However, exascale systems will spawn hundreds of thousands of threads in order to exploit their massive parallel architectures and thus conventional OS threads are too heavy for that purpose. Several lightweight thread (LWT) libraries have recently appeared offering lighter mechanisms to tackle massive concurrency. In order to examine the suitability of LWTs in high-level runtimes, we develop a set of microbenchmarks consisting of commonlyfound patterns in current parallel codes. Moreover, we study the semantics offered by some LWT libraries in order to expose the similarities between different LWT application programming interfaces. This study reveals that a reduced set of LWT functions can be sufficient to cover the common parallel code patterns and that those LWT libraries perform better than OS threads-based solutions in cases where task and nested parallelism are becoming more popular with new architectures.

Castello, Adrian↗

Collaborative Research: A Model of Partially Ionized Plasma Flows with Kinetic Treatment of Neutral Atoms and Nonthermal Ions

Interactions of flows of partially ionized, magnetized plasma are frequently accompanied by the presence of both thermal and non-thermal (pickup) ion components. Such interactions cannot be modeled using traditional MHD equations and require more advanced approaches to treat them. If a nonthermal component of ions is formed due to charge exchange and collisions between the thermal (core) ions and neutrals, it experiences the action of magnetic field, its distribution function is isotropized, and it soon acquires the velocity of the ambient plasma without being thermodynamically equilibrated. This situation, e. g., takes place in the outer heliosphere - the part of interstellar space beyond the solar system whose properties are determined by the solar wind interaction with the local interstellar medium. This is also possible in laboratory, at million degrees and above, when plasma is conducting electricity far too well, which makes Ohmic heating ineffective. To attain the target temperatures one needs additional heating eventually playing a dominant role. Among such sources is a so-called neutral particle beam heating. This is a wide-spread technique (Joint European Torus and International Thermonuclear Experimental Reactor experiments) based on the injection of powerful beams of neutral atoms into ohmically preheated plasma. In this project we have investigated the energy and density separation between the thermal and nonthermal components in the solar wind and interstellar plasmas. A new model has been developed in which we solve the ideal MHD equations for mixture of all ions and the kinetic Boltzmann equation to describe the transport of neutral atoms. As a separate capability, we can treat the flow of neutral atoms in a multi-component fashion, where neutral atoms born in each thermodynamically distinct regions are governed by the Euler gas dynamic equations. We also describe the behavior of pickup ions either kinetically, using the Fokker–Planck equation, or as a separate fluid. Our numerical simulations have demonstrated that pickup ions play a major role in the interaction of the solar wind and (partially ionized) interstellar medium plasmas. Our teams have investigated the stability of the surface (the heliopause) that separates the solar wind from the local interstellar medium, the transport of galactic cosmic rays, the properties of the heliotail flow, and modifications to the bow wave in front of the heliopause due to charge exchange between the neutral H atoms born in the solar wind and interstellar ions. Modeling results have been validated against observational data, such as obtained by the Interstellar Boundary Explorer (IBEX), and made it possible to shed light on the structure of energetic neutral atom maps created by this spacecraft.. We have also demonstrated that charge-exchange modulated heliosphere is a source of anisotropy of the multi-TeV cosmic ray flux observed in a number of Earth-bound air shower experiments. Newly developed codes are implemented within a Multi-Scale Fluid-Kinetic Simulation Suite (MS-FLUKSS), a publicly available code being developed by our team for over 12 years. MS-FLUKSS scales well up to 160,000 computing cores and has been ported on major supercomputers in the country. Efficient parallelization and data choreography in the continuum simulation modules are provided by Chombo, an adaptive mesh refinement framework managed by Phillip Colella’s team at LBNL. We have implemented in-house, hybrid (MPI+OpenMP) parallelization of the kinetic modules that solve the Boltzmann equation with a Monte Carlo method. Currently, the kinetic modules are being rewritten to take advantage of the modern CPU-GPU supercomputer architecture. The scope of the project allowed us to enhance plasma research and education in such broad, multidis- ciplinary field as physics of partially ionized plasma and its application to space physics and fusion science. Besides the impact on the modeling of complex physical systems, our approach to computational resource management for complex codes utilizing multiple algorithm technologies appears to be a major advance on current approaches. The development of sophisticated resource management will be essential for all future modeling efforts that incorporate a diversity of scales and physical processes. Our effort provided leadership in promoting computational science and plasma physics within the UAH and FIT campuses and, through the training of a broad spectrum of scientists and engineers, foster new technologies across the country.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Parallelized Kalman-Filter-Based Reconstruction of Particle Tracks on Many-Core Architectures

Faced with physical and energy density limitations on clock speed, contemporary microprocessor designers have increasingly turned to on-chip parallelism for performance gains. Algorithms should accordingly be designed with ample amounts of fine-grained parallelism if they are to realize the full performance of the hardware. This requirement can be challenging for algorithms that are naturally expressed as a sequence of small-matrix operations, such as the Kalman filter methods widely in use in high-energy physics experiments. In the High-Luminosity Large Hadron Collider (HL-LHC), for example, one of the dominant computational problems is expected to be finding and fitting charged-particle tracks during event reconstruction; today, the most common track-finding methods are those based on the Kalman filter. Experience at the LHC, both in the trigger and offline, has shown that these methods are robust and provide high physics performance. Previously we reported the significant parallel speedups that resulted from our efforts to adapt Kalman-filter-based tracking to many-core architectures such as Intel Xeon Phi. Here we report on how effectively those techniques can be applied to more realistic detector configurations and event complexity.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

GSoFa: Scalable Sparse Symbolic LU Factorization on GPUs

Decomposing a matrix $\mathbf {A}$ into a lower matrix $\mathbf {L}$ and an upper matrix $\mathbf {U}$, which is also known as LU decomposition, is an essential operation in numerical linear algebra. For a sparse matrix, LU decomposition often introduces more nonzero entries in the $\mathbf {L}$ and $\mathbf {U}$ factors than in the original matrix. A symbolic factorization step is needed to identify the nonzero structures of $\mathbf {L}$ and $\mathbf {U}$ matrices. Attracted by the enormous potentials of the Graphics Processing Units (GPUs), an array of efforts have surged to deploy various LU factorization steps except for the symbolic factorization, to the best of our knowledge, on GPUs. This article introduces gSoFa, the first GPU-based symbolic factorization design with the following three optimizations to enable scalable LU symbolic factorization for nonsymmetric pattern sparse matrices on GPUs. First, here we introduce a novel fine-grained parallel symbolic factorization algorithm that is well suited for the Single Instruction Multiple Thread (SIMT) architecture of GPUs. Second, we tailor supernode detection into a SIMT friendly process and strive to balance the workload, minimize the communication and saturate the GPU computing resources during supernode detection. Third, we introduce a three-pronged optimization to reduce the excessive space consumption problem faced by multi-source concurrent symbolic factorization. Taken together, gSoFa achieves up to 31× speedup from 1 to 44 Summit nodes (6 to 264 GPUs) and outperforms the state-of-the-art CPU project, on average, by 5×. Notably, gSoFa also achieves up to 47 percent of the peak memory throughput of a V100 GPU in the Summit Supercomputer.

97 MATHEMATICS AND COMPUTING↗