Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “KERNELS”

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 919 records · Page 51
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Towards On-Chip Learning for Low Latency Reasoning with End-to-End Synthesis

The Software Defined Architectures (SODA) Synthesizer is an open-source compiler-based tool able to automatically generate domain-specialized systems targeting Application-Specific Integrated Circuits (ASICs) or Field Programmable Gate Arrays (FPGAs) starting from high-level programming. SODA is composed of a frontend, SODA-OPT, which leverages the multilevel intermediate representation (MLIR) framework to interface with productive programming tools (e.g., machine learning frame-works), identify kernels suitable for acceleration, and perform high-level optimizations, and of a state-of-the-art high-level synthesis backend, Bambu from the PandA framework, to generate custom accelerators. One specific application of the SODA Synthesizer is the generation of accelerators to enable ultra-low latency inference and control on autonomous systems for scientific discovery (e.g., electron microscopes, sensors in particle accelerators, etc.). This paper provides an overview of the flow in the context of the generation of accelerators for edge processing to be integrated in transmission electron microscopy (TEM) devices, focusing on use cases from precision material synthesis. We show the tool in action with an example of design space exploration for inference on reconfigurable devices with a conventional deep neural network model (LeNet). Finally, we discuss the research directions and opportunities enabled by SODA in the area of autonomous control for scientific experimental workflows.

Castellana, Vito G.↗
DOE OSTI2023

Revisiting Temporal Blocking Stencil Optimizations

Iterative stencils are used widely across the spectrum of High Performance Computing (HPC) applications. Many efforts have been put into optimizing stencil GPU kernels, given the prevalence of GPU-accelerated supercomputers. To improve the data locality, temporal blocking is an optimization that combines a batch of time steps to process them together. Under the observation that GPUs are evolving to resemble CPUs in some aspects, we revisit temporal blocking optimizations for GPUs. We explore how temporal blocking schemes can be adapted to the new features in the recent Nvidia GPUs, including large scratchpad memory, hardware prefetching, and device-wide synchronization. We propose a novel temporal blocking method, EBISU, which champions low device occupancy to drive aggressive deep temporal blocking on large tiles that are executed tile-by-tile. We compare EBISU with state-of-the-art temporal blocking libraries: STENCILGEN and AN5D. We also compare with state-of-the-art stencil auto-tuning tools that are equipped with temporal blocking optimizations: ARTEMIS and DRSTENCIL. Over a wide range of stencil benchmarks, EBISU achieves speedups up to 2.53x and a geometric mean speedup of 1.49x over the best state-of-the-art performance in each stencil benchmark.

Zhang, Lingqi↗
DOE OSTI2023

Software-Hardware Co-design of Heterogeneous SmartNIC System for Recommendation Models Inference and Training

Deep Learning Recommendation Models (DLRMs) are critical applications in various domains and have evolved as one of the single largest machine learning applications. Trillions of DLRM parameters exceed the on-chip memory capacity of GPUs. Large-scale multi-node systems are required for distributed DLRM inference and training, which suffer from the all-to-all communication bottleneck, mainly limiting the scalability of ever-growing DLRMs. In recent years, SmartNICs have evolved with coupled computation and communication capabilities providing opportunities for a powerful heterogeneous device in the system. However, there isn't such a distributed system that fully leverages the abundant smartNIC resources that resolve the scalability issue of DLRMs. In this work, we proposed a software-hardware co-design of a heterogeneous smartNIC system that resolves the communication bottleneck of distributed DLRMs, mitigates the memory bandwidth pressure, and improves computation efficiency. We provide a set of smartNIC designs of cache systems (including local cache and remote cache) and smartNIC computation kernels which reduce data movement, relieve memory lookup intensity, and improve the GPU's computation efficiency. In addition, we propose a graph algorithm that improves the data locality of queries within batches which optimizes the overall system performance with higher data reuse. Our evaluation shows that our system achieves 2.1x latency speedup for inference and 1.6x throughput speedup for training.

Guo, Anqi↗
DOE OSTI2023

BitGNN: Unlocking the Performance Potential of Binary Graph Neural Networks on GPUs

Graph Neural Networks (GNNs) have shown compelling results in many graph-based learning tasks. They are, however, time-consuming. Recent work has shown a promising direction in improving GNN speed and shrinking the size — network binarization, which binarizes network values and operations. Prior work, however, mainly focused on algorithm designs, leaving it open on how to fully materialize the performance potential. This work fills the gap by proposing techniques to best map binary GNNs and their computations to fit the nature of bit manipulations, optimizations and algorithms to maximize BSpMM kernel efficiency, and solutions to other factors influencing the end-to-end time on GPUs. Results on real-world graphs show that the proposed techniques outperform state of-the-art binary GNN implementations by 21-67× with little accuracy loss.

Chen, Jou-An↗
DOE OSTI2023

Performance Analysis and Optimal Node-aware Communication for Enlarged Conjugate Gradient Methods

Krylov methods are a key way of solving large sparse linear systems of equations but suffer from poor strong scalability on distributed memory machines. Furthermore, this is due to high synchronization costs from large numbers of collective communication calls alongside a low computational workload. Enlarged Krylov methods address this issue by decreasing the total iterations to convergence, an artifact of splitting the initial residual and resulting in operations on block vectors. In this article, we present a performance study of an enlarged Krylov method, Enlarged Conjugate Gradients (ECG), noting the impact of block vectors on parallel performance at scale. Most notably, we observe the increased overhead of point-to-point communication as a result of denser messages in the sparse matrix-block vector multiplication kernel. Additionally, we present models to analyze expected performance of ECG, as well as motivate design decisions. Most importantly, we introduce a new point-to-point communication approach based on node-aware communication techniques that increases efficiency of the method at scale.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

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

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

Neff, Reece W.↗
DOE OSTI2023

Tiling Framework for Heterogeneous Computing of Matrix based Tiled Algorithms

Tiling matrix operations can improve the load balancing and performance of applications on heterogeneous computing resources. Writing a tile-based algorithm for each operation with a traditional, hand-tuned tiling approach that uses for loops in C/C++ is cumbersome and error prone. Moreover, it must enable and support the heterogeneous memory management of data objects and also explore architecture-supported, native, tiled-data transfer APIs instead of copying the tiled data to continuous memory before the data transfer. The tiling framework provides a tiled data structure for heterogeneous memory mapping and parameterization to a heterogeneous task specification API. We have integrated our tiled framework into MatRIS (Math kernels library using IRIS). IRIS is a heterogeneous run-time framework with a heterogeneous programming model, memory model, and task execution model. Experiments reveal that the tiled framework for BLAS operations has improved the programmability of tiled BLAS and improved performance by ~20% when compared against the traditional method that copies the data to continuous memory locations for heterogeneous computing.

Miniskar, Narasinga Rao↗
DOE OSTI2023

HetArch: Heterogeneous Microarchitectures for Superconducting Quantum Systems

Noisy Intermediate-Scale Quantum Computing (NISQ) has dominated headlines in recent years, with the longer-term vision of Fault-Tolerant Quantum Computation (FTQC) offering significant potential but at currently intractable resource costs and quantum error correction (QEC) overheads. For problems of interest, FTQC will require millions of physical qubits with long coherence times, high-fidelity gates, and compact sizes to surpass classical systems. Just as heterogeneous specialization has offered scaling benefits in classical computing, it is likewise gaining interest in FTQC. However, systematic use of heterogeneity in either hardware or software elements of FTQC systems remains a serious challenge due to the vast design space and the variable physical constraints. This paper meets the challenge of making heterogeneous FTQC design practical by introducing HetArch, a toolbox for designing heterogeneous quantum systems, and using it to explore heterogeneous design scenarios. Using a hierarchical approach, we successively break quantum algorithms into smaller operations (akin to classical application kernels), thus greatly simplifying the design space and resulting tradeoffs. Specializing to superconducting systems, we then design optimized heterogeneous hardware composed of varied superconducting devices, abstracting physical constraints into design rules that enable devices to be assembled into standard cells optimized for specific operations, which, in turn, form heterogeneous modules optimized for quantum subroutines. Finally, we provide a heterogeneous design space exploration framework which reduces the simulation burden by a factor of 10^4 or more and allows us to characterize optimal design points. We use these techniques to design superconducting quantum modules for entanglement distillation, error correction, and code teleportation, reducing error rates by 2.6×, 10.7×, and 3.4× compared to homogeneous systems.

Quantum Computing, Quantum Physics, Computer Archi↗
DOE OSTI2024

RAP: Resource-aware Automated GPU Sharing for Multi-GPU Recommendation Model Training and Input Preprocessing

Ensuring high-quality recommendations for newly onboarded users requires the continuous retraining of Deep Learning Recommendation Models (DLRMs) with freshly generated data. To serve the online DLRM retraining, existing solutions use hundreds of CPU computing nodes designated for input preprocessing, causing significant power consumption that surpasses even the power usage of GPU trainers. To this end, we propose RAP, an end-to-end DLRM training framework that supports Resource-aware Automated GPU sharing for DLRM input Preprocessing and Training. The core idea of RAP is to accurately capture the remaining GPU computing resources during DLRM training for input preprocessing, achieving superior training efficiency without requiring additional resources. Specifically, RAP utilizes a co-running cost model to efficiently assess the costs of various input preprocessing operations, and it implements a resource-aware horizontal fusion technique that adaptively merges smaller kernels according to GPU availability, circumventing any interference with DLRM training. In addition, RAP leverages a heuristic searching algorithm that jointly optimizes both the input preprocessing graph mapping and the co-running schedule to maximize the end-to-end DLRM training throughput. The comprehensive evaluation shows that RAP achieves 78.3× speedup on average over CPU-based DLRM input preprocessing frameworks. In addition, the end-to-end training throughput of RAP is only 2.04% lower than the ideal case, which has no input preprocessing overhead.

Wang, Zheng↗
DOE OSTI2023

FFTX-IRIS: Towards Performance Portability and Heterogeneity for SPIRAL Generated Code

FFTX-IRIS is a dynamic system to efficiently utilize novel heterogeneous platforms. This system links two next-generation frameworks, FFTX and IRIS, to navigate the complexity of different hardware architectures. FFTX provides a runtime code generation framework for high-performance Fast Fourier Transform kernels. IRIS runtime provides portability and multi-device heterogeneity, allowing computation on any available compute resource. Together, FFTX-IRIS enables code generation, seamless portability, and performance without user involvement. We show the design of the FFTX-IRIS system along with an evaluation of various small FFT benchmarks. We also demonstrate multi-device heterogeneity of FFTX-IRIS with a larger stencil application.

Rao, Sanil↗
DOE OSTI2023

Demonstrating Cross-Facility Data Processing at Scale with Laue Microdiffraction

In February and April 2023 live, at-scale data processing demonstrations were conducted between the Advanced Photon Source (APS), a synchrotron light source, and the Argonne Leadership Computing Facility (ALCF). These tests were run as part of a novel beamline technique: coded aperture laue micro-diffraction. This technique requires a significant amount of compute to decode appeture patterns embedded in the detector stream. An autonomous system was able to send data to ALCF during an experiment, utilize 50 nodes of the Polaris supercomputer to process 6-12 hour scans, and return the data back to the APS within 12-15 minutes behind the detector. With scan points arriving every 72 seconds, the system kept up with the beamline, potentially enabling in-experiment analysis. The data processing system utilizes Globus infrastructure and an on-demand queue to dynamically acquire nodes on Polaris. The underlying reconstruction algorithms were parallelized via MPI and accelerated with custom CUDA kernels.

Prince, Michael↗
DOE OSTI2023

cuAlign: Scalable Network Alignment on GPU Accelerators

Given two graphs, the objective of network alignment is to find the best one-to-one mapping of vertices in one graph (??) to vertices in the other (??), such that the number of overlaps is maximized. We say that edges(??, ??) ???and(??', ??') ??? are overlapped if ?? is mapped to ??' and ?? is mapped to??'. Network alignment is an important optimization problem with several applications in bioinformatics, computer vision and ontology matching. Since it is an NP-hard problem, efficient heuristics and scalable implementations are necessary. In this work, we introduce a new framework that combines the concepts of intra-network proximity using vertex embedding,Belief Propagation (BP) and approximate weighted matching, and provides qualitative improvements up to22%over state-of-the-art approaches. We also provide scalable implementations on GPU accelerators, demonstrating up to19×speedup for Belief Propagation and 3× speedup for approximate weighted matching relative to previous multithreaded implementation. A combination of combinatorial and algebraic kernels within the network alignment algorithm poses significant hurdles for parallelization. Load imbalance and irregular DRAM traffic limit achievable performance on GPUs. Our novel approach identifies and exploits unique structural proper-ties of the BP-based algorithm and employs code fusion to reduce data movement between different steps of the algorithm. Using a diverse set of inputs, we demonstrate qualitative improvements of our algorithms, and performance gains of our GPU-accelerated implementation. We believe that our work will enable algorithmic improvements and practical applications of network alignment.

Xiang, Lizhi↗
DOE OSTI2024

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↗
DOE OSTI2024

The Case for Co-Designing Model Architectures with Hardware

While GPUs are responsible for training the vast majority of state-of-the-art deep learning models, the implications of their architecture are often overlooked when designing new deep learning (DL) models. As a consequence, modifying a DL model to be more amenable to the target hardware can significantly improve the runtime performance of DL training and inference. In this paper, we provide a set of guidelines for users to maximize the runtime performance of their transformer models. These guidelines have been created by carefully considering the impact of various model hyperparameters controlling model shape on the efficiency of the underlying computation kernels executed on the GPU. We find the throughput of models with “efficient” model shapes is up to 39% higher while preserving accuracy compared to models with a similar number of parameters but with unoptimized shapes.

Yin, Junqi↗
DOE OSTI2024

SPARTA: High-Level Synthesis of Parallel Multi-Threaded Accelerators

This article presents a methodology for the Synthesis of PARallel multi-Threaded Accelerators (SPARTA) from OpenMP annotated C/C++ specifications. SPARTA extends an open-source HLS tool, enabling the generation of accelerators that provide latency tolerance for irregular memory accesses through multithreading, support fine-grained memory-level parallelism through a hot-potato deflection-based network-on-chip (NoC), support synchronization constructs, and can instantiate memory-side caches. Our approach is based on a custom runtime OpenMP library, providing flexibility and extensibility. Experimental results show high scalability when synthesizing irregular graph kernels. The accelerators generated with our approach are, on average, 2.29x faster than state-of-the-art HLS methodologies.

Design automation↗
DOE OSTI2025

Advancing Quantum Many-Body GW Calculations on Exascale Supercomputing Platforms

Advanced ab initio materials simulations face growing challenges as increasing systems and phenomena complexity requires higher accuracy, driving up computational demands. Quantum many-body GW methods are state-of-the-art for treating electronic excited states and couplings but often hindered due to the costly numerical complexity. Here, we present innovative implementations of advanced GW methods within the BerkeleyGW package, enabling large-scale simulations on Frontier and Aurora exascale platforms. Our approach demonstrates exceptional versatility for complex heterogeneous systems with up to 17,574 atoms, along with achieving true performance portability across GPU architectures. We demonstrate excellent strong and weak scaling to thousands of nodes, reaching double-precision core-kernel performance of 1.069 ExaFLOP/s on Frontier (9,408 nodes) and 707.52 PetaFLOP/s on Aurora (9,600 nodes), corresponding to 59.45% and 48.79% of peak, respectively. Our work demonstrates a breakthrough in utilizing exascale computing for quantum materials simulations, delivering unprecedented predictive capabilities for rational designs of future quantum technologies.

Zhang, Benran [University of Southern California, ↗
DOE OSTI2025

ScaWL: Scaling k-WL (Weisfeiler-Lehman) Algorithms in Memory and Performance on Shared and Distributed-Memory Systems

The k-dimensional Weisfeiler-Lehman (k-WL) algorithm—developed as an efficient heuristic for testing if two graphs are isomorphic—is a fundamental kernel for node embedding in the emerging field of graph neural networks. Unfortunately, the k-WL algorithm has exponential storage requirements, limiting the size of graphs that can be handled. This work presents a novel k-WL scheme with a storage requirement orders of magnitude lower while maintaining the same accuracy as the original k-WL algorithm. Due to the reduced storage requirement, our scheme allows for processing much bigger graphs than previously possible on a single compute node. For even bigger graphs, we provide the first distributed-memory implementation. Our k-WL scheme also has significantly reduced communication volume and offers high scalability. Our experimental results demonstrate that our approach is significantly faster and has superior scalability compared to five other implementations employing state-of-the-art techniques.

algorithims↗
DOE OSTI2025

Communication Lower Bounds and Optimal Algorithms for Symmetric Matrix Computations

In this article, we focus on the communication costs of three symmetric matrix computations: (i) multiplying a matrix with its transpose, known as a symmetric rank-k update (SYRK) (ii) adding the result of the multiplication of a matrix with the transpose of another matrix and the transpose of that result, known as a symmetric rank-2k update (SYR2K) (iii) performing matrix multiplication with a symmetric input matrix (SYMM). All three computations appear in the Level 3 Basic Linear Algebra Subroutines (BLAS) and have wide use in applications involving symmetric matrices. We establish communication lower bounds for these kernels using sequential and distributed-memory parallel computational models, and we show that our bounds are tight by presenting communication-optimal algorithms for each setting. Our lower bound proofs rely on applying a geometric inequality for symmetric computations and analytically solving constrained nonlinear optimization problems. As a result, the symmetric matrix and its corresponding computations are accessed and performed according to a triangular block partitioning scheme in the optimal algorithms.

Al Daas, Hussam [Rutherford Appleton Laboratory, D↗