Search NASA⌕ Search

SEARCH · Search NASA

Results for “program architectures”

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 739 records · Page 41

ARCUS Vertical-Axis Wind Turbine (Final Scientific/Technical Report)

While land-based wind energy has become economically competitive with traditional energy generation sources in the U.S., offshore wind is not. For floating offshore wind this difference is even more substantial where the levelized cost of energy (LCOE) is projected to be around 3-5 times more expensive than land-based wind. The turbine capital costs represent around 50% of the LCOE for land-based wind sites, but the increased system costs for floating offshore wind reduce this to 20%. The platform and mooring costs are the single largest contributor to the LCOE for floating offshore wind where their mass must counteract the overturning moment caused by the turbine’s thrust force. Despite the high costs of the platform and relatively low cost of the turbine, current offshore wind turbines are designed essentially the same as for land-based sites. Reducing the LCOE is the greatest challenge to realize the benefits of sustained development of floating offshore wind in the U.S. Reducing the complicated system costs of floating offshore wind will enable the industry to continue to grow and outpace current projections if reduced cost curves can be reached. The ideal wind energy system would remove all mass and cost that is not directly capturing energy from the wind. For floating offshore wind energy systems, this objective is even more significant as increased mass above the water level must be supported by larger and more expensive floating platforms. For this reason, vertical-axis wind turbines (VAWTs) are ideal for floating offshore sites and have several advantages over horizontal-axis wind turbines (HAWTs) at this scale. Large VAWTs offer opportunities for improved energy capture over HAWTs as single units and with reduced wind plant aerodynamic losses through enhanced wake recovery. Additionally, the platform-level placement of the VAWT drivetrain greatly reduces the demands placed on the floating platform and its mass and cost. The ARCUS vertical-axis wind turbine concept (U.S. 11,421,650 B2) is an iteration beyond traditional Darrieus-type VAWTs that replaces the rigid tower with blades that are bent into shape and held in place with tensioned center supports, like a bow. The ARCUS design has been shown to further decrease the VAWT rotor mass properties, with a 50% reduction over traditional Darrieus VAWTs quantified in the ATLANTIS program. The ARCUS VAWT’s efficient use of material for the rotor and turbine support structures combined with its lowered center of gravity enables a tension-leg platform (TLP) with simplified installation procedures. TLPs have been an emerging platform architecture in the Oil and Gas industry and demonstrated to have the lowest hull mass requirements while maintaining stability with minimal roll and pitch deflections in operation. A 22 MW ARCUS turbine has been designed with a three-column TLP that enables quayside integration of the turbine while maintaining system stability during tow-out and installation and having optimal mass and cost properties. A comprehensive analysis shows the optimal ARCUS TLP system design minimizes LCOE through efficient material usage and increased energy capture to yield a competitive LCOE estimate of $\$$55/MWh. A comparison with a reference HAWT, having the same swept area, quantifies the advantages that helped to produce this improved LCOE for the ARCUS concept: (1) 30% reduction in total turbine mass, (2) 70% reduction in turbine center of gravity, and (3) 45% increase in energy production over what is optimal for a HAWT. Intellectual property has been generated through the ATLANTIS program providing opportunities to further reduce the LCOE and improve the performance of the ARCUS turbine and TLP system, expanding the list of innovations to support commercial development of the ARCUS concept.

17 WIND ENERGY↗

Root genetics in the field to understand drought adaptation and carbon sequestration (Final Scientific/Technical Report)

For all crop plants, roots play a critical role in growth. Roots anchor the plants, and are the primary site of nutrient and water uptake. Roots are also the main source of C to soil in the form of root tissues and exudates, and thus greatly influence SOM stocks. To perform these functions, primary roots extend into soil, producing a network of branching roots of characteristic form, known as its root system architecture (RSA). RSA varies among species, and among varieties within a species that are adapted to different environments. Root traits are major targets for the second green revolution because of their potential to improve crop productivity, increase drought tolerance and nutrient acquisition, and increase C capture of soil. Improving the quality of roots in maize will be particularly valuable, since this crop is planted on over 92 million acres annually in the US. The future sustainability of agricultural systems relies on their ability to enhance soil organic matter (SOM) storage and reduce GHG emissions, while maintaining or enhancing productivity. This program had two components, Sensors and Models. For the first component, we designed and built a high-throughput phenotyping platform for root pulling of maize plants. This eliminated the physical labor of manually pulling up plants and reduced the number of personnel required down to one. The standardized pulling mechanism allowed recording force curves during the pulling process, providing additional information. We validated that the maximum force for pulling the root system was well-correlated with the root system mass and provided root crowns for further RSA analysis. These root crowns identified significant correlations with 2D root area and root depth, along with 3D root volume, total root length and number of root tips. We then used this system for field-based studies in maize on the genetics of root system architecture and its relation to nitrogen-use efficiency (NUE), including using lines relevant to the Corteva breeding program. Varieties were also evaluated at Corteva sites in the cornbelt and Danforth farm in Missouri, to establish responses across sites. From these studies we have identified genetic loci associated with root traits and created mutant lines for these loci and correlations of root traits with NUE. For the Models component, we worked to incorporate root and soil characteristics into the MEMS 2.0 soil and ecosystem biogeochemical model. Existing soil C models, such as Century, are unable to represent specific root trait interactions with the soil environment and therefore to accurately forecast the potential C sequestration benefits of root breeding under different climatic and soil type conditions. We have developed the MEMS 2.0 ecosystem biogeochemical model to improve quantification of farm-scale soil carbon and greenhouse gas emissions. The new knowledge and large datasets produced by this project will be used to develop and drive an innovative model capable of forecasting the impacts on soil C stocks and nutrient dynamics. An innovation was to use the empirical data from the field studies (in 1, above) to model genetic variation in nitrogen use efficiencies and soil C input. Our work demonstrated that maize root-derived C rapidly replaces existing soil C and after 3 years of continuous maize, up to 20% of soil organic C in the topsoil (0-15cm) and 3% in the subsoil (15-30cm) was contributed by maize. However, this contribution did not entirely represent a net increase. Root C contribution to soil was affected by maize genetics. We have analyzed soils derived from the CSU field trials for C and N stocks, in the different soil physical fractions represented by the MEMS model, using both physical fractionation with elemental analyses, and Fourier transformed infrared spectroscopy. Data will be used to link crop nitrogen use efficiencies with soil C sequestration and provide data to bridge the field trials with the model development, for verification of model predictions. The project had a number of successful outcomes: we have used the new phenotyping platform to identify new genetic loci that can enhance root phenotypes; we have partnered with multiple maize seed companies phenotype varieties in their breeding programs; we have developed the MEMS model that can help inform industry on the potential for carbon sequestration in the agricultural sector, and which is now available at the CSU Soil Carbon Solutions Center for use.

59 BASIC BIOLOGICAL SCIENCES↗

Hierarchical Distributed Optimal Power Flow of HV and MV Distribution Networks With Continuous and Discrete Devices

With large-scale distributed photovoltaics (PVs) being integrated into distribution networks (DNs), coordinated optimal power flow (OPF) of high voltage (HV) and medium voltage (MV) DNs should be investigated to optimally dispatch the distributed PVs and other network devices. Here, this paper presents a hierarchical distributed OPF method for HV and MV DNs with on-load tap changers, reactive power compensators, feeder switches and distributed PVs. A hierarchical master-slave control architecture is applied to implement coordinated OPF of two-layer DNs. The HV master problem and MV subproblems are transformed into mixed-integer convex problems respectively with second order cone programming and LinDistFlow approximation. Since there is no efficient distributed algorithm to solve such OPF models with integer subproblems, a novel distributed algorithm is proposed in this paper to efficiently solve the hierarchical coordinated OPF model with integer subproblems in a distributed manner. In the proposed algorithm, the coordinated OPF model is solved in a branch-and-bound framework, where in each branch node generalized Benders decomposition (GBD) algorithm is applied to decompose the coordinated OPF model into a master problem and relaxed subproblems and solves them iteratively to get optimal solution. The GBD optimal and feasible cutting planes generated in a branch node are proved to be valid for its descendants. Moreover, three acceleration techniques are introduced into the proposed algorithm to improve computational efficiency. Finally, the effectiveness and accuracy of the proposed method are verified via simulation tests in Jinzhai DNs of China.

42 ENGINEERING↗

Time-sliced quantum circuit partitioning for modular architectures

Current quantum computer designs will not scale. To scale beyond small prototypes, quantum architectures will likely adopt a modular approach with clusters of tightly connected quantum bits and sparser connections between clusters. We exploit this clustering and the statically-known control flow of quantum programs to create tractable partitioning heuristics which map quantum circuits to modular physical machines one time slice at a time. Specifically, we create optimized mappings for each time slice, accounting for the cost to move data from the previous time slice and using a tunable lookahead scheme to reduce the cost to move to future time slices. We compare our approach to a traditional statically-mapped, owner-computes model. Our results show strict improvement over the static mapping baseline. Here, we reduce the non-local communication overhead by 89.8% in the best case and by 60.9% on average. Our techniques, unlike many exact solver methods, are computationally tractable.

Baker, Jonathan M.↗

xesn: Echo state networks powered by Xarray and Dask

Xesn is a Python package that allows scientists to easily design Echo State Networks (ESNs) for forecasting problems. ESNs are a Recurrent Neural Network architecture introduced by Jaeger (2001) that are part of a class of techniques termed Reservoir Computing. One defining characteristic of these techniques is that all internal weights are determined by a handful of global, scalar parameters, thereby avoiding problems during backpropagation and reducing training time significantly. Because this architecture is conceptually simple, many scientists implement ESNs from scratch, leading to questions about computational performance. Xesn offers a straightforward, standard implementation of ESNs that operates efficiently on CPU and GPU hardware. The package leverages optimization tools to automate the parameter selection process, so that scientists can reduce the time finding a good architecture and focus on using ESNs for their domain application. Importantly, the package flexibly handles forecasting tasks for out-of-core, multi-dimensional datasets, eliminating the need to write parallel programming code. Xesn was initially developed to handle the problem of forecasting weather dynamics, and so it integrates naturally with Python packages that have become familiar to weather and climate scientists such as Xarray (Hoyer & Hamman, 2017). However, the software is ultimately general enough to be utilized in other domains where ESNs have been useful, such as in signal processing (Jaeger & Haas, 2004).

97 MATHEMATICS AND COMPUTING↗

Performance Engineering Research Institute SciDAC-2 Enabling Technologies Institute (Final Report)

Enhancing the performance of SciDAC applications on petascale systems had high priority within DOE SC at the start of the second phase of the SciDAC program, SciDAC-2, as it continues to do so today. Achieving expected levels of performance on high-end computing (HEC) systems is growing ever more challenging due to enormous scale, increasing architectural complexity, and increasing application complexity. To address these challenges, the University of Southern California's Information Sciences Institute organized the Performance Engineering Research Institute (PERI). PERI implemented a unified, tripartite research plan encompassing: (1) performance modeling and prediction; (2) automatic performance tuning; and (3) performance engineering of high profile applications. Within PERI, USC's primary research activity was automatic tuning (autotuning) of scientific software. This activity was spurred by the strong user preference for automatic tools and was based on previous successful activities such as ATLAS, which automatically tuned components of the LAPACK linear algebra library, and other recent work on autotuning domain-specific libraries. Our other major component was application engagement, to which we devoted approximately 30% of our effort to work directly with SciDAC-2 applications. This report is a summary of the overall results of the USC PERI effort.

97 MATHEMATICS AND COMPUTING↗

Summary Report on the SAE 2016 Range Extenders for Electric Vehicles Symposium

The SAE 2016 Range Extenders for Electric Vehicles Symposium was a 2-day technical meeting focused on the role of advanced internal combustion engines (ICEs) and other novel energy converter technologies for extending the range of electric vehicles (EVs). The first-of-its-kind symposium was notable for focusing solely on the range extender (REx) technologies and not the EVs. The technical program featured presentations from international leaders from industry, government, national laboratories, and academia. The opening keynote presentations covered a broad range of topics including consumer behavior, policy implications, regulatory considerations, and REx architectures as enablers for advanced technologies. The technical sessions focused on an array of REx technologies including conventional ICEs, as well as less conventional or emerging technologies such as microturbines, fuel cells, low-temperature combustion engines, and aluminum-air batteries. The symposium included two panel sessions. The trend toward increasing vehicle electrification and the changing role of ICEs and other auxiliary power unit technologies for use as REx’s is leading to new research and design development needs. The symposium captured the interest of the industry and research communities in exploring the opportunities and challenges associated with REx’s for EVs. This report includes key takeaways, summarized below, and draft notes for each presentation and panel discussion.

33 ADVANCED PROPULSION SYSTEMS↗

Integration of the Nek5000 Computational Fluid Dynamics Code with the NEAMS Workbench

The mission of the US Department of Energy’s Nuclear Energy Advanced Modeling and Simulation (NEAMS) Program is to develop, apply, deploy, and support state-of-the-art predictive modeling and simulation tools for the design and analysis of current and future nuclear energy systems using computing architectures ranging from laptops to leadership-class facilities. NEAMS is organized into three product lines: the Fuels Product Line (FPL),the Reactors Product Line (RPL), and the Integration Product Line (IPL). The FPL and RPL focus on the development of high-fidelity tools such as the MOOSE-based applications BISON and MARMOT [3, 9] at Idaho National Laboratory, and the PROTEUS neutronics code and the NEK5000 computational fluid dynamics code from the Simulation-based High-efficiency Advanced Reactor Prototyping (SHARP) framework [10]. The NEAMS IPL will enable users to apply high-fidelity simulations to inform lower order models for the design, analysis, and licensing of advanced nuclear systems.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

DEVELOPMENT AND ASSESSMENT OF COATINGS FOR FUTURE POWER GENERATION TURBINES

The NETL-Regional University Alliance (RUA) continues to advance technology development critical to turbine manufacturer efforts for achieving DOE Fossil Energy (FE's) Advanced Turbine Program Goals. In conjunction with NETL, Coatings for Industry (CFI), the University of Pittsburgh, NASA GRC, and Corrosion Control Inc., efforts have been focused on development of composite thermal barrier coating (TBC) architectures that consist of an extreme temperature coating, a commercially applied 7-8 YSZ TBC, a reduced cost bond coat, and a diffusion barrier coating that are applied to nickel-based superalloys or single crystal airfoil substrate materials for use at temperatures >1450 C (> 2640 F). Additionally, construction of a unique, high temperature ({approx}1100 C; {approx}2010 F), bench-scale, micro-indentation, nondestructive (NDE) test facility at West Virginia University (WVU) was completed to experimentally address in-situ changes in TBC stiffness during extended cyclic oxidation exposure of coated single crystal coupons in air or steam containing environments. The efforts and technical accomplishments in these areas are presented in the following sections of this paper.

36 MATERIALS SCIENCE↗

OpenCGRA: An Open-Source Unified Framework for Modeling,Testing, and Evaluating CGRAs

Coarse-grained reconfigurable arrays (CGRAs),loosely defined as arrays of functional units (e.g, adder, sub-tractor, multiplier, divider, or larger multi-operation units, butsmaller than a general-purpose core) interconnected through aNetwork-on-Chip, provide higher flexibility than domain-specificASIC accelerators while offering increased hardware efficiencywith respect to fine-grained reconfigurable devices, such as FieldProgrammable Gate Arrays (FPGAs). The fast evolving fieldsof machine learning and edge computing, which are seeing acontinuous flow of novel algorithms and larger models, makeCGRAs ideal target architectures to allow domain specializationwithout loosing too much generality. They also generally offerquicker and more effective reconfigurability than FPGAs, po-tentially allowing adaptation during actual algorithm execution,and implement a dataflow programming paradigm that adaptswell to these emerging workloads. Designing and generating aCGRA, however, still requires to define the type and number ofthe specific functional units, implement their interconnect andthe network topology, and perform its simulation and validation,given a variety of workloads of interest.In this paper, we propose OpenCGRA, a Python-based unifiedframework that integrates generation, modeling, testing and eval-uation for CGRAs. OpenCGRA is the first open-source integratedframework able to support the full top-to-bottom design flow forspecializing and implementing CGRAs: modeling at different ab-straction levels (functional level, cycle level, register-transfer level),generation, simulation, testing at different granularities (unit test-ing, integration testing, property-based testing), and characteriza-tion (area, power, and timing). OpenCGRAs will be made availableon GitHub.

CGRA, synthesis↗

An inertia-free filter line-search algorithm for large-scale nonlinear programming

We present a filter line-search algorithm that does not require inertia information of the linear system. This feature enables the use of a wide range of linear algebra strategies and libraries, which is essential to tackle large-scale problems on modern computing architectures. The proposed approach performs curvature tests along the search step to detect negative curvature and to trigger convexification. We prove that the approach is globally convergent and we implement the approach within a parallel interior-point framework to solve large-scale and highly nonlinear problems. Our numerical tests demonstrate that the inertia-free approach is as efficient as inertia detection via symmetric indefinite factorizations. We also demonstrate that the inertia-free approach can lead to reductions in solution time because it reduces the amount of convexification needed.

Chiang, Nai-Yuan↗

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

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

97 MATHEMATICS AND COMPUTING↗

INFORMATION: Management Alert on the Department's Monitoring of the Weatherization Assistance Program in the State of Illinois

Under the American Recovery and Reinvestment Act of 2009 (Recovery Act), the Department of Energy's Weatherization Assistance Program received $5 billion to improve the energy efficiency of homes owned or occupied by low income persons, reduce their total residential expenditures, and improve their health and safety. Since the Recovery Act was enacted in February 2009, the Department has awarded weatherization grants to every state, the District of Columbia and five territories. Because of the unprecedented level of funding and the risks associated with spending vast amounts of money in a relatively short period of time, the Office of Inspector General (OIG) initiated a series of audits designed to evaluate the Program's internal control structures at both the Federal and state levels. As part of our work, we are in the process of reviewing Weatherization Program internal controls for the State of Illinois. We are also currently performing identical audits in the States of North Carolina, Pennsylvania and Virginia. Under the Recovery Act and the Department's Program, the State of Illinois received $242 million to weatherize 26,933 homes. The State of Illinois awarded these funds to 35 local agencies responsible for determining recipients' eligibility, contracting for the installation of the weatherization work, and conducting final inspections to ensure that work on homes was done in accordance with requirements. Inspectors working for the local agencies are required to evaluate the quality of mechanical and architectural improvements, such as furnace installations and window caulking, and certify that the work performed meets established standards. Under a Department approved plan in place at the time of our review, state officials were required to evaluate the sufficiency of local agency monitoring controls and to inspect the work performed on at least five percent of the units weatherized with Department funds during the program year for each local agency. We identified significant internal control deficiencies in the management of the Weatherization Program in Illinois which require immediate attention. Specifically, our audit testing revealed significant problems with on-site monitoring and inspection of the Illinois Home Weatherization Assistance Program (Illinois). We noted that the Department had not fulfilled its requirement to perform monitoring visits at the State level. In addition, Illinois officials had not complied with the Department's requirements for inspecting weatherization work conducted by local agencies. Finally, we found that a weatherization inspection for one of the local agencies failed to detect substandard installation of energy saving materials. This case involved a furnace gas leak that could have resulted in serious injury to the occupants and material damage to the structure. This is an interim report and our audit work remains in progress.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

Programmable Site‐Specific Functionalization of DNA Origami with Polynucleotide Brushes

Abstract Combining surface‐initiated, TdT (terminal deoxynucleotidyl transferase) catalyzed enzymatic polymerization (SI‐TcEP) with precisely engineered DNA origami nanostructures (DONs) presents an innovative pathway for the generation of stable, polynucleotide brush‐functionalized DNA nanostructures. We demonstrate that SI‐TcEP can site‐specifically pattern DONs with brushes containing both natural and non‐natural nucleotides. The brush functionalization can be precisely controlled in terms of the location of initiation sites on the origami core and the brush height and composition. Coarse‐grained simulations predict the conformation of the brush‐functionalized DONs that agree well with the experimentally observed morphologies. We find that polynucleotide brush‐functionalization increases the nuclease resistance of DONs significantly, and that this stability can be spatially programmed through the site‐specific growth of polynucleotide brushes. The ability to site‐specifically decorate DONs with brushes of natural and non‐natural nucleotides provides access to a large range of functionalized DON architectures that would allow for further supramolecular assembly, and for potential applications in smart nanoscale delivery systems.

Yang, Yunqi↗

SYCL for Performance Portability: Application Experience with Coupled Cluster Formalism in Quantum Chemistry on Exascale Systems

The exascale computing has brought unprecedented heterogeneity in node architectures, with systems such as Frontier and Aurora featuring diverse GPU accelerators, network connectivity among others. Ensuring performance portability across these platforms is a key challenge. To address this, we employ the SYCL programming model to develop portable, high-performance quantum chemistry workloads. As a representative application, we focus on the non-iterative Triples component of the coupled-cluster CCSD(T) method, a key driver in quantum chemistry. In this work, we report on our experience deploying SYCL-based implementations using both DPC++ and AdaptiveCPP across two flagship exascale platforms: OLCF Frontier with AMD MI250X GPUs and ALCF Aurora with Intel GPUs. Our results demonstrate that SYCL enables efficient, single-source implementations that scale to thousands of nodes, delivering performance on par with vendor-optimized HIP solutions. We highlight key insights into runtime behavior, kernel portability, and scaling characteristics, showing that SYCL offers a viable path for performance-portable computing.

Bagusetty, Abhishek [Argonne National Laboratory (↗

A High Performance Sparse Tensor Algebra Compiler in MLIR

Sparse tensor algebra is widely used in many applications, including scientific computing, machine learning, and data analytics. The performance of sparse tensor algebra kernels strongly depends on the intrinsic characteristics of the input tensors, hence many storage formats are designed for tensors to achieve optimal performance for particular applications/architectures, which makes it challenging to implement and optimize every tensor operation of interest on a given architecture. We propose a tensor algebra domain-specific language (DSL) and compiler framework to automatically generate kernels for mixed sparse-dense tensor algebra operations. The proposed DSL provides high-level programming abstractions that resemble the familiar Einstein notation to represent tensor algebra operations. The compiler introduces a new Sparse Tensor Algebra dialect built on top of LLVM's extensible MLIR compiler infrastructure for efficient code generation while covering a wide range of tensor storage formats. Our compiler also leverages input-dependent code optimization to enhance data locality for better performance. Our results show that the performance of automatically generated kernels outperforms the state-of-the-art sparse tensor algebra compiler, with up to 20.92x, 6.39x, and 13.9x performance improvement over state-of-the-art tensor algebra compilers, for parallel SpMV, SpMM, and TTM, respectively.

Tian, Ruiqin↗

SmartFuse: Reconfigurable Smart Switches to Accelerate Fused Collectives in HPC Applications

Communication switches have sometimes been augmented to process collectives (e.g., the IBM BlueGene project and the Mellanox SHArP switch). In this work, we find that there is a great acceleration opportunity through the further augmentation of switches to accelerate more complex functions that combine communication with computation. We consider three types of such functions. The first is fully-fused collectives built by fusing multiple existing collectives like Allreduce with Alltoall. The second is semi-fused collectives built by combining a collective with another computation. The third we refer to as higher-order collectives built by combining multiple computations and communications, such as to perform matrix-matrix multiply (PGEMM). In this work, we propose a framework called SmartFuse to accelerate fused collective functions. The core of SmartFuse is a reconfigurable smart switch to support these operations. The semi/fully fused collectives are implemented with a CGRAlike architecture, while higher-order collectives are implemented with a more specialized computational unit that can also schedule communication. Supporting our framework is software to evaluate and translate relevant parts of the input program, compile them into a control data flow graph, and then map this graph to the switch hardware. The proposed framework, once deployed, has the strong potential to accelerate existing HPC applications transparently by encapsulation within an MPI implementation. Experimental results show that this approach improves the performance of the PGEMM kernel, MINIFE, and AMG by, on average, 94%, 15%, and 13%, respectively.

Haghi, Pouya↗