SEARCH · Search NASA

Results for “parallel, scaling”

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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,495 records · Page 83

All sources NASA NTRS DOE OSTI

DOE OSTI2024

QRCODE: Massively parallelized real-time time-dependent density functional theory for periodic systems

We present a new software module, QRCODE (Quantum Research for Calculating Optically Driven Excitations), for massively parallelized real-time time-dependent density functional theory (RT-TDDFT) calculations of periodic systems in the open-source Qbox software package. Our approach utilizes a custom implementation of a fast Fourier transformation scheme that significantly reduces inter-node message passing interface (MPI) communication of the major computational kernel and shows impressive scaling up to 16,344 CPU cores. In addition to improving computational performance, QRCODE contains a suite of various time propagators for accurate RT-TDDFT calculations. As benchmark applications of QRCODE, we calculate the current density and optical absorption spectra of hexagonal boron nitride (h-BN) and photo-driven reaction dynamics of the ozone-oxygen reaction. We also calculate the second and higher harmonic generation of monolayer and multi-layer boron nitride structures as examples of large material systems. Our optimized implementation of RT-TDDFT in QRCODE enables large-scale calculations of real-time electron dynamics of chemical and material systems with enhanced computational performance and impressive scaling across several thousand CPU cores.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2017

A Dynamical and Statistical Characterization of U.S. Extreme Precipitation Events and Their Associated Large-Scale Meteorological Patterns

Abstract Regional patterns of extreme precipitation events occurring over the continental United States are identified via hierarchical cluster analysis of observed daily precipitation for the period 1950–2005. Six canonical extreme precipitation patterns (EPPs) are isolated for the boreal warm season and five for the cool season. The large-scale meteorological pattern (LMP) inducing each EPP is identified and used to create a “base function” for evaluating a climate model’s potential for accurately representing the different patterns of precipitation extremes. A parallel analysis of the Community Climate System Model, version 4 (CCSM4), reveals that the CCSM4 successfully captures the main U.S. EPPs for both the warm and cool seasons, albeit with varying degrees of accuracy. The model’s skill in simulating each EPP tends to be positively correlated with its capability in representing the associated LMP. Model bias in the occurrence frequency of a governing LMP is directly related to the frequency bias in the corresponding EPP. In addition, however, discrepancies are found between the CCSM4’s representation of LMPs and EPPs over regions such as the western United States and Midwest, where topographic precipitation influences and organized convection are prominent, respectively. In these cases, the model representation of finer-scale physical processes appears to be at least equally important compared to the LMPs in driving the occurrence of extreme precipitation.

54 ENVIRONMENTAL SCIENCES↗

NASA NTRS1998

New Parallel Algorithms for Structural Analysis and Design of Aerospace Structures

Subspace and Lanczos iterations have been developed, well documented, and widely accepted as efficient methods for obtaining p-lowest eigen-pair solutions of large-scale, practical engineering problems. The focus of this paper is to incorporate recent developments in vectorized sparse technologies in conjunction with Subspace and Lanczos iterative algorithms for computational enhancements. Numerical performance, in terms of accuracy and efficiency of the proposed sparse strategies for Subspace and Lanczos algorithm, is demonstrated by solving for the lowest frequencies and mode shapes of structural problems on the IBM-R6000/590 and SunSparc 20 workstations.

Nguyen, Duc T.↗

DOE OSTI2016

Extreme-scale Algorithms and Solver Resilience

A widening gap exists between the peak performance of high-performance computers and the performance achieved by complex applications running on these platforms. Over the next decade, extreme-scale systems will present major new challenges to algorithm development that could amplify this mismatch in such a way that it prevents the productive use of future DOE Leadership computers due to the following; Extreme levels of parallelism due to multicore processors; An increase in system fault rates requiring algorithms to be resilient beyond just checkpoint/restart; Complex memory hierarchies and costly data movement in both energy and performance; Heterogeneous system architectures (mixing CPUs, GPUs, etc.); and Conflicting goals of performance, resilience, and power requirements.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

Dynamic Ride-Matching for Large-Scale Transportation Systems

Efficient dynamic ride-matching (DRM) in large-scale transportation systems is a key driver in transport simulations to yield answers to challenging problems. Although the DRM problem is simple to solve, it quickly becomes a computationally challenging problem in large-scale transportation system simulations. Therefore, this study thoroughly examines the DRM problem dynamics and proposes an optimization-based solution framework to solve the problem efficiently. To benefit from parallel computing and reduce computational times, the problem’s network is divided into clusters utilizing a commonly used unsupervised machine learning algorithm along with a linear programming model. Then, these sub-problems are solved using another linear program to finalize the ride-matching. At the clustering level, the framework allows users adjusting cluster sizes to balance the trade-off between the computational time savings and the solution quality deviation. A case study in the Chicago Metropolitan Area, U.S., illustrates that the framework can reduce the average computational time by 58% at the cost of increasing the average pick up time by 26% compared with a system optimum, that is, non-clustered, approach. Another case study in a relatively small city, Bloomington, Illinois, U.S., shows that the framework provides quite similar results to the system-optimum approach in approximately 62% less computational time.

33 ADVANCED PROPULSION SYSTEMS↗

DOE OSTI2024

Sparse Linear Solvers for Large-scale Electromagnetic Transient Simulations

Linear solvers form the basis for electromagnetic transient (EMT) simulations. There is a need to speed up EMT simulations as larger regions are analyzed using EMT simulations. For the same, the performance of linear solvers plays an important role. Exploiting the sparsity of the matrices generated in EMT simulations could assist with speed-up. Scalability is also crucial as power grids expand, demanding solutions capable of accommodating the increasing system size. Recent studies from the North American Electric Reliability Corporation (NERC) increasingly emphasize that EMT simulation models of the power grid will grow larger with the inclusion of power electronics components. Parallelisms in sparsity patterns exploit modern central processing units (CPUs), multi-core CPUs, and graphics processing units (GPUs) architectures in sparse solver designs. Therefore, this paper explores publicly available existing linear solvers and investigates their efficiency in large-scale power grid simulations. A large-scale power grid is developed by increasing the size of the IEEE 39 bus test system to up to 39000 bus systems.

Hsu, Kuan-Chieh↗

DOE OSTI2018

Evolution of A-Type Macrosegregation in Large Size Steel Ingot After Multistep Forging and Heat Treatment

A-type macrosegregation refers to the channel chemical heterogeneities that can be formed during solidification in large size steel ingots. In this research, a combination of experiment and simulation was used to study the influence of open die forging parameters on the evolution of A-type macrosegregation patterns during a multistep forging of a 40 metric ton (MT) cast, high-strength steel ingot. Macrosegregation patterns were determined experimentally by macroetch along the longitudinal axis of the forged and heat-treated ingot. Mass spectroscopy, on more than 900 samples, was used to determine the chemical composition map of the entire longitudinal sectioned surface. FORGE NxT 1.1 finite element modeling code was used to predict the effect of forging sequences on the morphology evolution of A-type macrosegregation patterns. For this purpose, grain flow variables were defined and implemented in a large scale finite element modeling code to describe oriented grains and A-type segregation patterns. Examination of the A-type macrosegregation showed four to five parallel continuous channels located nearly symmetrical to the axis of the forged ingot. In some regions, the A-type patterns became curved or obtained a wavy form in contrast to their straight shape in the as-cast state. Mass spectrometry analysis of the main alloying elements (C, Mn, Ni, Cr, Mo, Cu, P, and S) revealed that carbon, manganese, and chromium were the most segregated alloying elements in A-type macrosegregation patterns. The observed differences were analyzed using thermodynamic calculations, which indicated that changes in the chemical composition of the liquid metal can affect the primary solidification mode and the segregation intensity of the alloying elements. Finite element modeling simulation results showed very good agreement with the experimental observations, thereby allowing for the quantification of the influence of temperature and deformation on the evolution of the shape of the macrosegregation channels during the open die forging process.

36 MATERIALS SCIENCE↗

DOE OSTI2022

Scalable and accurate multi-GPU-based image reconstruction of large-scale ptychography data

Abstract While the advances in synchrotron light sources, together with the development of focusing optics and detectors, allow nanoscale ptychographic imaging of materials and biological specimens, the corresponding experiments can yield terabyte-scale volumes of data that can impose a heavy burden on the computing platform. Although graphics processing units (GPUs) provide high performance for such large-scale ptychography datasets, a single GPU is typically insufficient for analysis and reconstruction. Several works have considered leveraging multiple GPUs to accelerate the ptychographic reconstruction. However, most of these works utilize only the Message Passing Interface to handle the communications between GPUs. This approach poses inefficiency for a hardware configuration that has multiple GPUs in a single node, especially while reconstructing a single large projection, since it provides no optimizations to handle the heterogeneous GPU interconnections containing both low-speed (e.g., PCIe) and high-speed links (e.g., NVLink). In this paper, we provide an optimized intranode multi-GPU implementation that can efficiently solve large-scale ptychographic reconstruction problems. We focus on the maximum likelihood reconstruction problem using a conjugate gradient (CG) method for the solution and propose a novel hybrid parallelization model to address the performance bottlenecks in the CG solver. Accordingly, we have developed a tool, called PtyGer ( Pty chographic G PU(multipl e )-based r econstruction), implementing our hybrid parallelization model design. A comprehensive evaluation verifies that PtyGer can fully preserve the original algorithm’s accuracy while achieving outstanding intranode GPU scalability.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2017

A backward Monte-Carlo method for time-dependent runaway electron simulations

Kinetic descriptions of runaway electrons (REs) are usually based on Fokker-Planck models that determine the probability distribution function of REs in 2-dimensional momentum space. Despite the simplification involved, the Fokker-Planck equation can rarely be solved analytically and direct numerical approaches [e.g., continuum and particle-based Monte Carlo (MC)] can be time consuming, especially in the computation of asymptotic-type observables including the runaway probability, the slowing-down and runaway mean times, and the energy limit probability. In this study, we present a novel backward MC approach to these problems based on backward stochastic differential equations that describe the dynamics of the runaway probability by means of the Feynman-Kac theory. The key ingredient of the backward MC algorithm is to place all the particles in a runaway state and simulate them backward from the terminal time to the initial time. As such, our approach can provide much faster convergence than direct MC methods (by significantly reducing the number of particles required to achieve a prescribed accuracy) while at the same time maintaining the advantages of particle-based methods (compared to continuum approaches). The proposed algorithm is unconditionally stable and can be parallelized as easy as the direct MC method, and its extension to dimensions higher than two is straightforward, thus paving the way for conducting large-scale RE simulation.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

DOE OSTI2023

Porting fragmentation methods to GPUs using an OpenMP API: Offloading the resolution-of-the-identity second-order Møller–Plesset perturbation method

Here, using an OpenMP Application Programming Interface, the resolution-of-the-identity second-order Møller–Plesset perturbation (RI-MP2) method has been off-loaded onto graphical processing units (GPUs), both as a standalone method in the GAMESS electronic structure program and as an electron correlation energy component in the effective fragment molecular orbital (EFMO) framework. First, a new scheme has been proposed to maximize data digestion on GPUs that subsequently linearizes data transfer from central processing units (CPUs) to GPUs. Second, the GAMESS Fortran code has been interfaced with GPU numerical libraries (e.g., NVIDIA cuBLAS and cuSOLVER) for efficient matrix operations (e.g., matrix multiplication, matrix decomposition, and matrix inversion). The standalone GPU RI-MP2 code shows an increasing speedup of up to 7.5× using one NVIDIA V100 GPU with one IBM 42-core P9 CPU for calculations on fullerenes of increasing size from 40 to 260 carbon atoms using the 6-31G(d)/cc-pVDZ-RI basis sets. A single Summit node with six V100s can compute the RI-MP2 correlation energy of a cluster of 175 water molecules using the correlation consistent basis sets cc-pVDZ/cc-pVDZ-RI containing 4375 atomic orbitals and 14 700 auxiliary basis functions in ~0.85 h. In the EFMO framework, the GPU RI-MP2 component shows near linear scaling for a large number of V100s when computing the energy of an 1800-atom mesoporous silica nanoparticle in a bath of 4000 water molecules. The parallel efficiencies of the GPU RI-MP2 component with 2304 and 4608 V100s are 98.0% and 96.1%, respectively.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2019

Simulations of tokamak boundary plasma turbulence transport in setting the divertor heat flux width

The BOUT++ code has been used to simulate edge plasma electromagnetic (EM) turbulence and transport, and to study the role of EM turbulence in setting the scrape-off layer (SOL) heat flux width λ q . More than a dozen tokamak discharges from C-Mod, DIII-D, EAST, ITER and CFETR have been simulated with encouraging success. The parallel electron heat fluxes onto the target from the BOUT++ simulations of C-Mod, DIII-D, and EAST follow the experimental heat flux width scaling of the inverse dependence on the poloidal magnetic field. Further turbulence statistics analysis shows that the blobs are generated near the pedestal pressure peak gradient region inside the separatrix and contribute to the transport of the particle and heat in the SOL region. Transport simulations indicate two distinct regimes: drift dominant regime and turbulence dominant regime. Goldston's heuristic drift-based (HD) model yields a consistent divertor heat flux width in the drift dominant regime. For C-Mod enhanced D α H-mode discharges, drifts and turbulence are competing in setting the divertor heat flux width, possibly due to its compact machine size and good pedestal confinement. The simulations for ITER and CFETR indicate that divertor heat flux width λ q of the future machines may no longer follows the 1/B pol,OMP HD-based empirical (Eich) scalings and the HD model gives a pessimistic limit of divertor heat flux width. The simulation results show a transition from a drift dominant regime to a turbulence dominant regime from current machines to future machines such as ITER and CFETR for two reasons. (1) The magnetic drift-based radial transport decreases due to large CFETR and ITER machine sizes. (2) The SOL turbulence thermal diffusivity increases due to larger turbulent fluxes ejected from the pedestal into the SOL when operating in a different pedestal structure, from an ELM-free H-mode pedestal regime to a small and grassy ELM regime.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

DOE OSTI2019

UniverseMachine: The correlation between galaxy growth and dark matter halo assembly from z = 0−10

ABSTRACT We present a method to flexibly and self-consistently determine individual galaxies’ star formation rates (SFRs) from their host haloes’ potential well depths, assembly histories, and redshifts. The method is constrained by galaxies’ observed stellar mass functions, SFRs (specific and cosmic), quenched fractions, ultraviolet (UV) luminosity functions, UV–stellar mass relations, IRX–UV relations, auto- and cross-correlation functions (including quenched and star-forming subsamples), and quenching dependence on environment; each observable is reproduced over the full redshift range available, up to 0 < z < 10. Key findings include the following: galaxy assembly correlates strongly with halo assembly; quenching correlates strongly with halo mass; quenched fractions at fixed halo mass decrease with increasing redshift; massive quenched galaxies reside in higher-mass haloes than star-forming galaxies at fixed galaxy mass; star-forming and quenched galaxies’ star formation histories at fixed mass differ most at z < 0.5; satellites have large scatter in quenching time-scales after infall, and have modestly higher quenched fractions than central galaxies; Planck cosmologies result in up to 0.3 dex lower stellar – halo mass ratios at early times; and, none the less, stellar mass–halo mass ratios rise at z > 5. Also presented are revised stellar mass – halo mass relations for all, quenched, star-forming, central, and satellite galaxies; the dependence of star formation histories on halo mass, stellar mass, and galaxy SSFR; quenched fractions and quenching time-scale distributions for satellites; and predictions for higher-redshift galaxy correlation functions and weak lensing surface densities. The public data release (DR1) includes the massively parallel (>105 cores) implementation (the UniverseMachine), the newly compiled and remeasured observational data, derived galaxy formation constraints, and mock catalogues including lightcones.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2024

Operational experience and R&D results using the Google Cloud for High-Energy Physics in the ATLAS experiment

The ATLAS experiment at CERN relies on a Worldwide Distributed Computing Grid infrastructure to support its physics program at the Large Hadron Collider. ATLAS has integrated cloud computing resources to complement its Grid infrastructure and conducted an R&D program on Google Cloud Platform. These initiatives leverage key features of commercial cloud providers: lightweight configuration and operation, elasticity and availability of diverse infrastructures. Here this paper examines the seamless integration of cloud computing services as a conventional Grid site within the ATLAS workflow management and data management systems, while also offering new setups for interactive, parallel analysis. It underscores pivotal results that enhance the on-site computing model and outlines several R&D projects that have benefited from large-scale, elastic resource provisioning models. Furthermore, this study discusses the impact of cloud-enabled R&D projects in three domains: accelerators and AI/ML, ARM CPUs and columnar data analysis techniques.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

DOE OSTI2024

FuseIM: Fusing Probabilistic Traversals for Influence Maximization on Exascale Systems

Probabilistic breadth-first traversals (BPTs) are used in many network science and graph machine learning applications. In this paper, we are motivated by the application of BPTs in stochastic diffusion-based graph problems such as influence maximization. These applications heavily rely on BPTs to implement a Monte-Carlo sampling step for their approximations. Given the large sampling complexity, stochasticity of the diffusion process, and the inherent irregularity in real-world graph topologies, efficiently parallelizing these BPTs remains significantly challenging. In this paper, we present a new algorithm to fuse massive number of concurrently executing BPTs with random starts on the input graph. Our algorithm is designed to fuse BPTs by combining separate traversals into a unified frontier on distributed multi-GPU systems. To show the general applicability of the fused BPT technique, we have incorporated it into two state-of-the-art influence maximization parallel implementations (gIM and Ripples). Our experiments on up to 4K nodes of the OLCF Frontier supercomputer (32,768 GPUs and 196K CPU cores) show strong scaling behavior, and that fused BPTs can improve the performance of these implementations up to 34x (for gIM) and ~360x (for Ripples).

Neff, Reece W.↗

DOE OSTI2024

IDAES-PSE 2.5.0 Release

The Institute for the Design of Advanced Energy Systems (IDAES) Integrated Platform is a versatile computational environment offering extensive process systems engineering (PSE) capabilities for optimizing the design and operation of complex, interacting technologies and systems. IDAES enables users to efficiently search vast, complex design spaces to discover the lowest cost, most environmentally sustainable solutions while supporting the full process modeling lifecycle, from conceptual design to dynamic optimization and control. The extensible, open platform empowers users to create models of novel processes and rapidly develop custom analyses, workflows, and end-user applications. IDAES-PSE 2.5.0 Release Highlights Upcoming Changes IDAES will be switching to the new Pyomo solver interface in the next release. Whilst this will hopefully be a smooth transition for most users, there are a few important changes to be aware of. The new solver interface uses a different version of the IPOPT writer (“ipopt_v2”) and thus any custom configuration options you might have set for IPOPT will not carry over and will need to be reset. By default, the new Pyomo linear presolver will be activated with ipopt_v2. Whilst are working to identify any bugs in the presolver, it is possible that some edge cases will remain. IDAES will begin deploying a new set of scaling tools and APIs over the next few releases that make use of the new solver writers. The old scaling tools and APIs will remain for backward compatibility but will begin to be deprecated. New Models, Tools and Features New diagnostics check for near-parallel variables and constraints. New diagnostics tools for identifying causes of infeasibility in models. New example for creating a custom model of a liquid-liquid extractor unit operation. Bug Fixes Fixed bug in Gibbs reactor that caused it to appear to have additional spurious degrees of freedom. Fixed bug in the Modular Property Framework that would cause errors when trying to use phase-based material balances with phase equilibria. Fixed bug in Modular Properties Framework that caused errors when initializing models with non-vapor-liquid phase equilibria. Testing and Robustness Deployed the IDAES Diagnostics Toolbox to confirm that there are no structural or numerical issues in the core model libraries. Additional robustness tests for core model, and some associated improvements in the converge tester class. Fixed a number of issues that were causing unexpected warnings to be emitted during testing. Deprecations and Removals Removed examples for RIPE tool which has not been supported for a number of releases.

AS↗

DOE OSTI2017

Improvement of Alstom's Limestone-Based Chemical Looping Combustion Process for Higher Purity Flue Gas Production

Alstom Power Inc., a wholly owned subsidiary of the General Electric Company (GE), is developing Limestone-based Chemical Looping (LCL™) Processes, utilizing limestone derived calcium sulfate as an oxygen carrier to transport oxygen from air to the fuel. The LCL™ process is a breakthrough technology with the potential to capture CO 2 and SO 2 from new and existing coal-fired power plants at reduced costs while maintaining high plant power generation efficiency. Chemical looping is a next generation advancement to oxy-combustion which would not require the significant investment in a cryogenic distillation type oxygen production plant. Instead of the energy-consuming oxygen plant, chemical looping uses a benign material to directly remove oxygen from the air and bring that oxygen into contact with the fuel. This "game changing" technology is the lowest cost-of-electricity technical approach that GE Power has identified to date for coal power with carbon capture. The technology has significantly lower capital costs compared to a conventional steam power plant with first generation carbon capture technologies. This new power plant concept is based on a hybrid combustion-gasification process utilizing high temperature chemical and thermal looping technology. In the air reactor, a solid oxygen carrier picks up oxygen from air through an oxidation reaction and leaves nitrogen behind. The hot oxygen-carrying solid carrier is then transported to a fuel reactor, where it releases oxygen. The released oxygen then converts the coal in the fuel reactor into combustion gases. The solid carrier is recycled back to the air reactor for regeneration completing the “chemical loop.” The regeneration step produces heat from which steam is produced for power generation while the process results in a concentrated carbon dioxide stream in which the carbon is captured for use or sequestration (CCUS). Limestone as a precursor for the oxygen carrier is abundant, benign, low cost, and is used in CFB and PC boilers for sulfur control. Challenges associated with this carrier include a complex reaction mechanism with possible side reactions leading to the release of SO 2 . The present project (DOE/NETL Cooperative Agreement No. DE-FE0025073) report focuses on coal-fired Limestone Chemical Looping Combustion (LCL-C™) processes that can produce a higher purity flue gas stream and achieve an improved overall performance that will meet the stated DOE goal of generating power from coal while achieving greater than 90% CO2 removal at less than 35% increase in the Levelized Cost of Electricity (LCOE). It builds on the concurrent and synergistic limestone-based chemical looping combustion (LCL-C™) and gasification (LCL-G™) projects with the Department of Energy, that were executed in parallel with the present project. In this report, the LCL-C™ experimental testing from the electrically-heated, once-through, 100 kW Pilot-Scale Test Facility (PSTF) is analysed and reviewed. The PSTF Oxidizer is a 50-foot tall, 100-mm diameter transport reactor, while the PSTF Reducer reactor is based upon a bubbling-bed design. The experimental data from designated test periods (denoted “Weeks”) are reviewed and discussed. The application of high sulfur release in the Reducer (with proper sulfur management) exceeded the LCL-CTM performance targets. Furthermore, the ilmenite blend came within 1 percentage point of meeting all of the performance targets.

01 COAL, LIGNITE, AND PEAT↗

DOE OSTI2018

Evaluation of statistical variation of microstructural properties and temperature effects on creep fracture of Grade 91

This report describes three improvements made to a physically-based model for creep and creep rupture in Grade 91 steel developed as part of the Advanced Reactor Technologies program: 1. Progress on transitioning the model framework to the MOOSE finite element package to improve parallel scalability, increase the physical size of the simulations, and incorporate multiphysics effects into the model. 2. The development of a physically-based method of scaling the model parameters to accurately capture creep in Grade 91 for temperatures in the range of 450° to 500° C. 3. Extending the model to account for microstructural statistical variations, specifically capturing the effect of grain boundary energy on the physical parameters underling the model for grain boundary void nucleation and cavitation. Put together, these improvements result in an accurate, predictive model for the physical response of Grade 91 steel over the expected use temperature range for the material in future liquid metal cooled fast reactors. The model can be used to more accurately predict engineering properties of the material for very long service lives, which could lead to safer, more economical future advanced reactors.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

DOE OSTI2020

CDFG Extraction Tool for LLVM

With the dawn of the exascale era, computer scientists and engineers are faced with tremendous challenges across all facets of the HPC system - scalability, performance, reliability, and power consumption. In particular, the power-performance benefit from one processor generation to the next is seeing ever-diminishing returns and will require fundamental changes in the way we approach computation. In fact, it is likely that different applications will require different types of accelerators in order to meet power, performance, and reliability requirements at scale. One potential type of accelerator, a dataflow architecture, diverges from the traditional sequentially executed instruction model into one that reflects the inherent instruction-level parallelism in a program. This work presents the initial steps toward a tool that can extract the control-dataflow graph from an application.

97 MATHEMATICS AND COMPUTING↗