Search NASA⌕ Search

SEARCH · Search NASA

Results for “hardware accelerators”

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 19 records

Modeling and Predicting Performance of High Performance Computing Applications on Hardware Accelerators

Computers with hardware accelerators, also referred to as hybrid-core systems, speedup applications by offloading certain compute operations that can run faster on accelerators. Thus, it is not surprising that many of top500 supercomputers use accelerators. However, in addition to procurement cost, significant programming and porting effort is required to realize the potential benefit of such accelerators. Hence, before building such a system it is prudent to answer the question 'what is the projected performance benefit from accelerators for the workloads of interest?'. We address this question by way of a performance-modeling framework that predicts realizable application performance on accelerators rapidly and accurately without going to the considerable effort of porting and tuning. The modeling framework first automatically identifies commonly found compute patterns in scientific applications which we term idioms, which may benefit by accelerator technology. Next the framework models the predicted speedup of those idioms if they were to be ported to and run on hardware accelerators. As a proof of concept we characterize two kinds of accelerators 1) the FPGA accelerators on a Convey HC-1 system and 2) an NVIDIA FERMI GPU accelerator. We model performance of the idioms gather/scatter and stream and our predictions show that where these occur in two full-scale HPC applications, Milc and HYCOM, gather/scatter speeds up by as much as 15X, and stream by as much as 14X, whereas the overall compute time of Milc improves by 3.4% and HYCOM by 20%.

Computer Science↗

Generating clock signals for a cycle accurate, cycle reproducible FPGA based hardware accelerator

A method, system and computer program product are disclosed for generating clock signals for a cycle accurate FPGA based hardware accelerator used to simulate operations of a device-under-test (DUT). In one embodiment, the DUT includes multiple device clocks generating multiple device clock signals at multiple frequencies and at a defined frequency ratio; and the FPG hardware accelerator includes multiple accelerator clocks generating multiple accelerator clock signals to operate the FPGA hardware accelerator to simulate the operations of the DUT. In one embodiment, operations of the DUT are mapped to the FPGA hardware accelerator, and the accelerator clock signals are generated at multiple frequencies and at the defined frequency ratio of the frequencies of the multiple device clocks, to maintain cycle accuracy between the DUT and the FPGA hardware accelerator. In an embodiment, the FPGA hardware accelerator may be used to control the frequencies of the multiple device clocks.

97 MATHEMATICS AND COMPUTING↗

Generating clock signals for a cycle accurate, cycle reproducible FPGA based hardware accelerator

A method, system and computer program product are disclosed for generating clock signals for a cycle accurate FPGA based hardware accelerator used to simulate operations of a device-under-test (DUT). In one embodiment, the DUT includes multiple device clocks generating multiple device clock signals at multiple frequencies and at a defined frequency ratio; and the FPG hardware accelerator includes multiple accelerator clocks generating multiple accelerator clock signals to operate the FPGA hardware accelerator to simulate the operations of the DUT. In one embodiment, operations of the DUT are mapped to the FPGA hardware accelerator, and the accelerator clock signals are generated at multiple frequencies and at the defined frequency ratio of the frequencies of the multiple device clocks, to maintain cycle accuracy between the DUT and the FPGA hardware accelerator. In an embodiment, the FPGA hardware accelerator may be used to control the frequencies of the multiple device clocks.

Asaad, Sameh W.↗

Generating clock signals for a cycle accurate, cycle reproducible FPGA based hardware accelerator

A method, system and computer program product are disclosed for generating clock signals for a cycle accurate FPGA based hardware accelerator used to simulate operations of a device-under-test (DUT). In one embodiment, the DUT includes multiple device clocks generating multiple device clock signals at multiple frequencies and at a defined frequency ratio; and the FPG hardware accelerator includes multiple accelerator clocks generating multiple accelerator clock signals to operate the FPGA hardware accelerator to simulate the operations of the DUT. In one embodiment, operations of the DUT are mapped to the FPGA hardware accelerator, and the accelerator clock signals are generated at multiple frequencies and at the defined frequency ratio of the frequencies of the multiple device clocks, to maintain cycle accuracy between the DUT and the FPGA hardware accelerator. In an embodiment, the FPGA hardware accelerator may be used to control the frequencies of the multiple device clocks.

Asaad, Sameh W.↗

Even Higher-Level Synthesis: An Exploration of AI Hardware Accelerators using HLS4ML

With the rise of artificial intelligence, the popularization of deep learning, and a constantly evolving industry, the demand for flexible and efficient tools has never been greater. As algorithms grow more complex, their runtime and energy consumption increase exponentially. Customized hardware accelerators, long used for specific mathematical operations, remain essential for managing modern applications' computational and power demands. Hardware accelerators can speed up complex computations by orders of magnitude, but their manual design and verification processes are often challenging and time-consuming. High-Level Synthesis (HLS) provides a solution by transforming high-level algorithm descriptions, typically written in C++ or SystemC, into synthesizable RTL suitable for hardware implementation. This approach reduces development time for RTL engineers while offering flexibility beyond what traditional handwritten RTL can provide. We extended this capability to the machine-learning domain with the open-source framework hls4ml, which allows neural networks trained in Python frameworks like Tensorflow or PyTorch to be synthesized into efficient hardware representations for the traditional FPGA and ASIC flows. This breakthrough addresses the growing need for reduced design turnaround and easy verification of ML hardware accelerators with low latency and power efficiency constraints. During this tutorial, we will demonstrate how Python complements HLS by simplifying the ML design process, bridging the gap between software and hardware development. Attendees will explore how we translate neural networks modeled in Python into fixed-point C++ models suitable for HLS workflows. We will dive into strategies like Value-Range Analysis and Quantization-Aware Training, which optimize these designs for deployment and evaluate their accuracy, power consumption, and energy efficiency. To exemplify these concepts, experts from Fermilab will share their experiences applying this technology to high-energy physics experiments, where real-time, low-latency processing is critical. Over the years, Fermilab engineers have demonstrated how deep neural networks, optimized for hardware using hls4ml, can meet the stringent requirements of trigger systems at the CERN Large Hadron Collider. These systems rely on rapid decision-making to process immense data volumes while retaining only the most relevant events for further analysis. The application of hls4ml has also been extended to innovative technologies like smart pixel arrays. These smart pixels integrate ML inference capabilities directly into sensor devices, enabling localized data processing at the pixel level. This approach drastically reduces the need to transmit raw data to external processing units, significantly decreasing power consumption and latency. By embedding neural networks within the pixel architecture, the smart pixels can identify and prioritize relevant data in real time, providing a highly efficient solution for edge computing in scenarios such as particle detectors and imaging systems. Fermilab's work highlights the potential of hardware-accelerated ML in scenarios where both speed and power efficiency are mission-critical. Through this tutorial, attendees will gain valuable insights into the challenges and solutions of deploying ML in hardware. Understanding how HLS and hls4ml streamline the development of neural network-based hardware accelerators is fundamental for the industry's future. Participants will learn how these technologies are shaping the future of AI and scientific computing.

Di Guglielmo, Giuseppe [Fermilab]↗

An MLIR-based Compiler Flow for System-Level Design and Hardware Acceleration

The generation of custom hardware accelerators for applications implemented within high-level productive programming frameworks requires considerable manual effort. To automate this process, we introduce \sodaopt, a compiler tool that extends the MLIR infrastructure. \sodaopt automatically searches, outlines, tiles, and pre-optimizes relevant code regions to generate high-quality accelerators through high-level synthesis. \sodaopt can support any high-level programming framework and domain-specific language that interface with the MLIR infrastructure. By leveraging MLIR, \sodaopt solves compiler optimization problems with specialized abstractions. Backend synthesis tools connect to \sodaopt through progressive intermediate representation lowerings. \sodaopt interfaces to a design space exploration engine to identify the combination of compiler optimization passes and options that provides high-performance generated designs for different backends and targets. We demonstrate the practical applicability of the compilation flow by exploring the automatic generation of accelerators for deep neural networks operators outlined at arbitrary granularity and by combining outlining with tiling on large convolution layers. Experimental results with kernels from the PolyBench benchmark show that \sodaopt high-level optimizations improve execution delays of synthesized accelerators up to 60x. We also show that for the selected kernels, our solution outperforms the current of state-of-the art in more than 70% of the benchmarks and provides better average speedup in 55% of them.

Bohm Agostini, Nicolas↗

Hardware acceleration for HPS algorithms in two and three dimensions

We provide a flexible, open-source framework for hardware acceleration, namely massively-parallel execution on general-purpose graphics processing units (GPUs), applied to the hierarchical Poincaré–Steklov (HPS) family of algorithms for building fast direct solvers for linear elliptic partial differential equations. To take full advantage of the power of hardware acceleration, we propose two variants of HPS algorithms to improve performance on two- and three-dimensional problems. In the two-dimensional setting, we introduce a novel recomputation strategy that minimizes costly data transfers to and from the GPU; in three dimensions, we modify and extend the adaptive discretization technique of Geldermans and Gillman [1] to greatly reduce peak memory usage. We provide an open-source implementation of these methods written in JAX, a high-level accelerated linear algebra package, which allows for the first integration of a high-order fast direct solver with automatic differentiation tools. We conclude with extensive numerical examples showing our methods are fast and accurate on two- and three-dimensional problems.

Fast direct solvers↗

An FPGA-based hardware accelerator supporting sensitive sequence homology filtering with profile hidden Markov models

Abstract Background Sequence alignment lies at the heart of genome sequence annotation. While the BLAST suite of alignment tools has long held an important role in alignment-based sequence database search, greater sensitivity is achieved through the use of profile hidden Markov models (pHMMs). Here, we describe an FPGA hardware accelerator, called HAVAC, that targets a key bottleneck step (SSV) in the analysis pipeline of the popular pHMM alignment tool, HMMER. Results The HAVAC kernel calculates the SSV matrix at 1739 GCUPS on a $$\sim$$ ∼ $3000 Xilinx Alveo U50 FPGA accelerator card, $$\sim$$ ∼ 227× faster than the optimized SSV implementation in nhmmer . Accounting for PCI-e data transfer data processing, HAVAC is 65× faster than nhmmer’s SSV with one thread and 35× faster than nhmmer with four threads, and uses $$\sim$$ ∼ 31% the energy of a traditional high end Intel CPU. Conclusions HAVAC demonstrates the potential offered by FPGA hardware accelerators to produce dramatic speed gains in sequence annotation and related bioinformatics applications. Because these computations are performed on a co-processor, the host CPU remains free to simultaneously compute other aspects of the analysis pipeline.

59 BASIC BIOLOGICAL SCIENCES↗

Hardware Acceleration of Adaptive Neural Algorithms.

As tradit ional numerical computing has faced challenges, researchers have turned towards alternative computing approaches to reduce power - per - computation metrics and improve algorithm performance. Here, we describe an approach towards non - conventional computing that strengthens the connection between machine learning and neuroscience concepts. The Hardware Acceleration of Adaptive Neural Algorithms (HAANA) project ha s develop ed neural machine learning algorithms and hardware for applications in image processing and cybersecurity. While machine learning methods are effective at extracting relevant features from many types of data, the effectiveness of these algorithms degrades when subjected to real - world conditions. Our team has generated novel neural - inspired approa ches to improve the resiliency and adaptability of machine learning algorithms. In addition, we have also designed and fabricated hardware architectures and microelectronic devices specifically tuned towards the training and inference operations of neural - inspired algorithms. Finally, our multi - scale simulation framework allows us to assess the impact of microelectronic device properties on algorithm performance.

97 MATHEMATICS AND COMPUTING↗

hPIC2: A hardware-accelerated, hybrid particle-in-cell code for dynamic plasma-material interactions

The exascale era of high performance computing promises to bring the field of computational plasma physics ever closer to the goal of accurate multiscale modeling. Such computers will rely on hardware acceleration to offload work to dedicated components, notably general-purpose graphics processing units (GPUs). However, devices from different manufacturers require software to be written with different parallel programming models, greatly increasing the code maintenance burden of applications designed to perform on more than one such device. hPIC2 is a hybrid plasma simulation code developed with the Kokkos performance portability framework to target the architectures that will drive exascale computing for the foreseeable future. As a hybrid simulation code, hPIC2 investigates the simultaneous use of various plasma models on the same domain, at the same time. hPIC2 also optionally couples to RustBCA, which accurately models ion-material interactions using the binary collision approximation (BCA) method. In conclusion, hPIC2 therefore achieves scalable performance on a variety of computing architectures when simulating complex and diverse plasmas, particularly near plasma-material interfaces.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Map reduce using coordination namespace hardware acceleration

A system and method for supporting data MapReduce operations in a tuple space/coordinated namespace (CNS) extended memory storage architecture. The system-wide CNS provides an efficient means for storing and communicating data generated by local processes running at the nodes, and coordinated to provide MapReduce operations in a multi-nodal system. A hardware accelerated mechanism supports map reduce sorting/shuffle operations and reduce operations according to an aggregate function. Local processes running at a node generate a tuple corresponding to data generated by a process, each tuple having a tuple name and tuple data value corresponding to the generated data. Each tuple is processed and stored at the node or another node, dependent upon its tuple name. Tuple records associated with a tuple name are accumulated at one or more nodes according to a linked list structure at each that is accessible via a hash table index pointer at the node.

Jacob, Philip↗

Hardware-accelerated inference for real-time gravitational-wave astronomy

The field of transient astronomy has seen a revolution with the first gravitational-wave detections and the arrival of multi-messenger observations they enabled. Transformed by the first detection of binary black hole and binary neutron star mergers, computational demands in gravitational-wave astronomy are expected to grow by at least a factor of two over the next five years as the global network of kilometer-scale interferometers are brought to design sensitivity. With the increase in detector sensitivity, real-time delivery of gravitational-wave alerts will become increasingly important as an enabler of multi-messenger followup. In this work, we report a novel implementation and deployment of deep learning inference for real-time gravitational-wave data denoising and astrophysical source identification. This is accomplished using a generic Inference-as-a-Service model that is capable of adapting to the future needs of gravitational-wave data analysis. Overall, our implementation allows seamless incorporation of hardware accelerators and also enables the use of commercial or private (dedicated) as-a-service computing. Based on our results, we propose a paradigm shift in low-latency and offline computing in gravitational-wave astronomy. Such a shift can address key challenges in peak-usage, scalability and reliability, and provide a data analysis platform particularly optimized for deep learning applications. The achieved sub-millisecond scale latency will also be relevant for any machine learning-based real-time control systems that may be invoked in the operation of near-future and next generation ground-based laser interferometers, as well as the front-end collection, distribution and processing of data from such instruments.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Porting the Nonlinear Optimization Library HiOp to Accelerator-Based Hardware Architectures

While interior point method has been the centerpiece of nonlinear programming tools used in science and engineering, its reliance on linear solvers that can tackle sparse symmetric indefinite and highly ill-conditioned problems made it difficult to implement it effectively on hardware accelerators. HiOp optimization package attempts to provide an implementation of the interior point method suitable for hardware accelerators by compressing the original sparse problem to produce an underlying linear problem that is dense and of manageable size. Implementations of dense linear solvers are more mature and utilize hardware accelerators better than their sparse counterparts. There is a number of important domain problems, such as optimal power flow analysis for power grids, where the sparse problem can be effectively compressed and deploying dense linear solver within the interior point method can improve performance. Here we describe a portable implementation of HiOp optimization engine, which uses a linear solver from Magma library and runs entirely on hardware accelerators. To compress the problem, HiOp uses customized mixed dense-sparse linear algebra. All HiOp kernels are implemented using Umpire and RAJA portability libraries. We describe details of the implementation and discuss trade-offs between performance, portability and development cost.

97 MATHEMATICS AND COMPUTING↗

Batched matrix computations on hardware accelerators based on GPUs

Scientific applications require solvers that work on many small size problems that are independent from each other. At the same time, the high-end hardware evolves rapidly and becomes ever more throughput-oriented and thus there is an increasing need for an effective approach to develop energy-efficient, high-performance codes for these small matrix problems that we call batched factorizations. The many applications that need this functionality could especially benefit from the use of GPUs, which currently are four to five times more energy efficient than multicore CPUs on important scientific workloads. This study, consequently, describes the development of the most common, one-sided factorizations, Cholesky, LU, and QR, for a set of small dense matrices. The algorithms we present together with their implementations are, by design, inherently parallel. In particular, our approach is based on representing the process as a sequence of batched BLAS routines that are executed entirely on a GPU. Importantly, this is unlike the LAPACK and the hybrid MAGMA factorization algorithms that work under drastically different assumptions of hardware design and efficiency of execution of the various computational kernels involved in the implementation. Thus, our approach is more efficient than what works for a combination of multicore CPUs and GPUs for the problems sizes of interest of the application use cases. The paradigm where upon a single chip (a GPU or a CPU) factorizes a single problem at a time is not at all efficient in our applications’ context. We illustrate all of these claims through a detailed performance analysis. With the help of profiling and tracing tools, we guide our development of batched factorizations to achieve up to two-fold speedup and three-fold better energy efficiency as compared against our highly optimized batched CPU implementations based on MKL library. Finally, the tested system featured two sockets of Intel Sandy Bridge CPUs and we compared with a batched LU factorizations featured in the CUBLAS library for GPUs, we achieve as high as 2.5× speedup on the NVIDIA K40 GPU.

97 MATHEMATICS AND COMPUTING↗

Hardware-Accelerated Ray Tracing of CAD-Based Geometry for Monte Carlo Radiation Transport

Monte Carlo radiation transport (MCRT) methods have been used to simulate radiation environments for many decades by tracking individual particles through a model to accumulate statistical information. MCRT geometry is historically formed using the constructive solid geometry (CSG). Recently, significant work has been performed to support simulations using computer-aided design (CAD)-based tessellated surfaces to support highly complex geometries. Ray tracing acceleration data structures from the rendering and visualization community are applied to accelerate particle tracking in CAD-based models. Despite these efforts, CSG representations provide the superior performance in surface intersection operations during particle flight. Concurrently, pseudo Monte Carlo methods have become prevalent in rendering applications to support more realistic models for scattering media, motivating innovations that are advantageous for MCRT simulations. Finally, the authors’ work extends these innovations by employing Intel’s Embree ray tracing kernel within a geometry toolkit for Monte Carlo to improve the simulation performance using CAD-based models by factors of 1.5 to 2.

42 ENGINEERING↗

Power Grid Computational Challenges and Metrics for Hardware Accelerator Evaluation

The objective of this document is to support the DMC initiative’s research on hardware and software development. Our focus is on the power grid domain that has a rich diversity of numerical and computationally challenging applications. Through this document, we highlight two core building blocks, power flow, and transient stability equations, for power grid applications. Improving computational robustness and acceleration of these core blocks will tremendously benefit all power grid applications. We also highlight the most computationally dominant numerical kernels that are being used in power flow and transient stability.

24 POWER TRANSMISSION AND DISTRIBUTION↗