Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “computation offloading”

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

Method and apparatus for offloading compute resources to a flash co-processing appliance

Solid-State Drive (SSD) burst buffer nodes are interposed into a parallel supercomputing cluster to enable fast burst checkpoint of cluster memory to or from nearby interconnected solid-state storage with asynchronous migration between the burst buffer nodes and slower more distant disk storage. The SSD nodes also perform tasks offloaded from the compute nodes or associated with the checkpoint data. For example, the data for the next job is preloaded in the SSD node and very fast uploaded to the respective compute node just before the next job starts. During a job, the SSD nodes perform fast visualization and statistical analysis upon the checkpoint data. The SSD nodes can also perform data reduction and encryption of the checkpoint data.

97 MATHEMATICS AND COMPUTING↗

Computational Offload with BlueField Smart NICs

The recent introduction of a new generation of "smart NICs" have provided new accelerator platforms that include CPU cores or reconfigurable fabric in addition to traditional networking hardware and packet offloading capabilities. While there are currently several proposals for using these smartNICs for low-latency, in-line packet processing operations, there remains a gap in knowledge as to how they might be used as computational accelerators for traditional high-performance applications. This work aims to look at benchmarks and mini-applications to evaluate possible benefits of using a smartNIC as a compute accelerator for HPC applications. We investigate NVIDIA's current-generation BlueField-2 card, which includes eight Arm CPUs along with a small amount of storage, and we test the networking and data movement performance of these cards compared to a standard Intel server host. We then detail how two different applications, YASK and miniMD can be modified to make more efficient use of the BlueField-2 device with a focus on overlapping computation and communication for operations like neighbor building and halo exchanges. Our results show that while the overall compute performance of these devices is limited, using them with a modified miniMD algorithm allows for potential speedups of 5 to 20% over the host CPU baseline with no loss in simulation accuracy.

97 MATHEMATICS AND COMPUTING↗

Simulating Diverse HEP Workflows on Heterogeneous Architectures

We evaluate key patterns and estimate throughput bounds of simulated transformation of conventional high energy physics (HEP) data processing workflows to heterogeneous equivalents. The simulation parameter space includes the number of offloaded tasks, CPU/accelerator ratios of intra-task computations, offload latencies, and run time efficiency of offloaded computations. The simulation is performed for a diverse set of state-of-the-art event reconstruction scenarios from ATLAS, LHCb and CMS - the frontier HEP experiments of the Large Hadron Collider project.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Exploring OpenSNAPI Use Cases and Evolving Requirements [Slides]

Emerging system architectures are rapidly transforming in order to meet shifting requirements. Motivated by expanding data volumes, energy efficiency concerns, and the omnipresent need to improve performance, architectures are increasingly adopting a data-centric approach. At the core of this concept is the goal of minimizing data motion and instead processing data in-situ to the greatest degree possible. Therefore, data-centric designs, in contrast to conventional CPU-centric models, typically distribute compute capabilities throughout the architecture. As part of this paradigm shift, a novel class of devices known as data processing units (DPUs), alongside CPUs and GPUs, are quickly forming a third pillar of data-centric systems. These devices, which include smart network adapters and switches, seek to offload computation on data at the network edge as well as in-flight within the network fabric. The Open Smart Network API (OpenSNAPI) project seeks to develop a unified API for DPU devices. In our previous talks, we introduced the OpenSNAPI project and detailed our investigations regarding the viability of offloading compute intensive kernels to BlueField DPUs. In contrast, in this talk we detail our efforts to offload application-level file I/O to the DPU. We also discuss plans and early efforts to explore in-network compute capabilities. Finally, we describe our observations with respect to the evolving design of OpenSNAPI.

97 MATHEMATICS AND COMPUTING↗

How to program 122,400 heterogeneous cores and retain your sanity

Current technology trends favor hybrid architectures, typically with each node in a cluster containing both general-purpose and specialized 'accelerator' processors. The typical model for programming such systems is host-centric: The general-purpose processor orchestrates the computation, offloading performance-critical work to the accelerator, and data is communicated only among general-purpose processors. In this talk we propose a radically different hybrid-programming approach, which we call the 'reverse-acceleration model'. In this model the accelerators orchestrate the computation, offloading unacceleratable work to the general-purpose processors. Data is communicated among accelerators, not among general-purpose processors. We present the Cell Messaging Layer (CML), an implementation of the reverse-acceleration model for Los Alamos National Laboratory's Roadrunner supercomputer, a complex conglomerate of 122,400 processor cores of various types, multiple memory domains, and multiple network types, all with radically different performance characteristics but which together make Roadrunner the world's second-fastest supercomputer. CML demonstrates a new messaging-layer implementation technique called 'receiver-initiated message passing', which reduces communication latency by up to a third. Our thesis is that the reverse-acceleration model simplifies porting codes to heterogeneous systems and facilitates performance optimization. We present a case study of a legacy neutron-transport code that we modified to use reverse acceleration. Performance results from running this code across the full Roadrunner system indicate a substantial performance improvement over the unaccelerated version of the code.

97 MATHEMATICS AND COMPUTING↗

An identity privacy scheme for blockchain‐based on edge computing

Abstract Blockchain has decentralization characteristics and requires more targeted security schemes to protect user privacy. In contrast, existing signature schemes have many high‐complexity operations and impose an enormous computational burden on wireless nodes. This article proposes a light‐weighted identity privacy scheme for blockchain‐based on edge computing. We construct linkable identity privacy and non‐linkable identity privacy, which can resist collusion attacks while virtually guaranteeing blockchain nodes' identity privacy. Since edge computing offloads heavily, the proposed scheme has lower computational complexity than the existing techniques.

Mu, Rui↗

Combining In-situ and In-transit Processing to Enable Extreme-Scale Scientific Analysis

With the onset of extreme-scale computing, I/O constraints make it increasingly difficult for scientists to save a sufficient amount of raw simulation data to persistent storage. One potential solution is to change the data analysis pipeline from a post-process centric to a concurrent approach based on either in-situ or in-transit processing. In this context computations are considered in-situ if they utilize the primary compute resources, while in-transit processing refers to offloading computations toa set of secondary resources using asynchronous data transfers. In this paper we explore the design and implementation of three common analysis techniques typically performed on large-scale scientific simulations: topological analysis, descriptive statistics, and visualization. We summarize algorithmic developments, describe a resource scheduling system to coordinate the execution ofvarious analysis workflows, and discuss our implementation using the DataSpaces and ADIOS frameworks that support efficient data movement between in-situ and in-transit computations. Wedemonstrate the efficiency of our lightweight, flexible framework by deploying it on the Jaguar XK6 to analyze data generated by S3D, a massively parallel turbulent combustion code. Our framework allows scientists dealing with the data deluge at extreme scale to perform analyses at increased temporal resolutions, mitigate I/O costs, and significantly improve the time to insight.

Bennett, Janine C.↗

Multichannel meta-imagers for accelerating machine vision

Rapid developments in machine vision technology have impacted a variety of applications, such as medical devices and autonomous driving systems. These achievements, however, typically necessitate digital neural networks with the downside of heavy computational requirements and consequent high energy consumption. As a result, real-time decision-making is hindered when computational resources are not readily accessible. Here we report a meta-imager designed to work together with a digital back end to offload computationally expensive convolution operations into high-speed, low-power optics. Further, in this architecture, metasurfaces enable both angle and polarization multiplexing to create multiple information channels that perform positively and negatively valued convolution operations in a single shot. We use our meta-imager for object classification, achieving 98.6% accuracy in handwritten digits and 88.8% accuracy in fashion images. Owing to its compactness, high speed and low power consumption, our approach could find a wide range of applications in artificial intelligence and machine vision applications.

47 OTHER INSTRUMENTATION↗

Inference-Engine v0.1.0

Given a pre-trained neural network, Inference-Engine performs maps network inputs to outputs by executing the forward pass through the provided network. Although the predominant programming language for machine-learning is Python, most high-performance computing (HPC) applications are written in Fortran, C, or C++. Inference-Engine aims to support HPC programs and is written in Fortran, a language with a large feature set supporting interoperability with C. This software exposes concurrency in a portable way by using standard language features that some modern Fortran compilers can exploit with various optimizations, including offloading computation to a Graphics Processing Unit (GPU). In particular, this software makes extensive use of Fortran's "do concurrent" parallel loop construct, implicitly parallel array statements, and pure procedures that can be invoked inside "do concurrent" blocks. Inference-Engine also supports dynamic choice of inference methods at runtime. Two current options include one method that uses Fortran's "dot_product" intrinsic function inside "do concurrent" blocks and another method that instead uses Fortran' "matmul" array intrinsic function. We plan to investigate automatic compiler offloading of "do concurrent" calculations to GPUs and compile-time substitution of optimized libraries such as the Basic Linear Algebra Library (BLAS) for "matmul" invocations. We also envision the potential for the choice of which method to use could happen at program launch based on in situ performance measurements on any given platform.

Rouson, Damian↗

Targeting GPUs with OpenMP directives on Summit: A simple and effective Fortran experience

We use OpenMP to target hardware accelerators (GPUs) on Summit, a newly deployed supercomputer at the Oak Ridge Leadership Computing Facility (OLCF), demonstrating simplified access to GPU devices for users of our astrophysics code GenASiS and useful speedup on a sample fluid dynamics problem. We modify our workhorse class for data storage to include members and methods that significantly streamline the persistent allocation of and association to GPU memory. Users offload computational kernels with OpenMP target directives that are rather similar to constructs already familiar from multi-core parallelization. In this initial example we ask, “With a given number of Summit nodes, how fast can we compute with and without GPUs?”, and find total wall time speedups of ~ 12X. We also find reasonable weak scaling up to 8000 GPUs (1334 Summit nodes). We make available the source code from this work at https://github.com/GenASiS/GenASiS_Basics.

97 MATHEMATICS AND COMPUTING↗

Toward Evaluating High-Level Synthesis Portability and Performance between Intel and Xilinx FPGAs

Offloading computation from a CPU to a hardware accelerator is becoming a more common solution for improving performance because traditional gains enabled by Moore’s law and Dennard scaling have slowed. GPUs are often used as hardware accelerators, but field-programmable gate arrays (FPGAs) are gaining traction. FPGAs are beneficial because they allow hardware specific to a particular application to be created. However, they are notoriously difficult to program. To this end, two of the main FPGA manufacturers, Intel and Xilinx, have created tools and frameworks that enable the use of higher level languages to design FPGA hardware. Although Xilinx kernels can be designed by using C/C++, both Intel and Xilinx support the use of OpenCL C to architect FPGA hardware. However, not much is known about the portability and performance between these two device families other than the fact that it is theoretically possible to synthesize a kernel meant for Intel to Xilinx and vice versa.In this work, we evaluate the portability and performance of Intel and Xilinx kernels. We use OpenCL C implementations of a subset of the Rodinia benchmarking suite that were designed for an Intel FPGA and make the necessary modifications to create synthesizable OpenCL C kernels for a Xilinx FPGA. We find that the difficulty of porting certain kernel optimizations varies, depending on the construct. Once the minimum amount of modifications is made to create synthesizable hardware for the Xilinx platform, more nontrivial work is needed to improve performance. However, we find that constructs that are known to be performant for an FPGA should improve performance regardless of the platform; the difficulty comes in deciding how to invoke certain kernel optimizations while also abiding by the constraints enforced by a given platform’s hardware compiler.

Cabrera, Anthony↗

Accelerated Constrained Sparse Tensor Factorization on Massively Parallel Architectures

This study presents the first constrained sparse tensor factorization (cSTF) framework that optimizes and fully offloads computation to massively parallel GPU architectures, and the first performance characterization of cSTF on GPU architectures. In contrast to prior work on tensor factorization, where the matricized tensor times Khatri-Rao product (MTTKRP) is the primary performance bottleneck, our systematic analysis of the cSTF algorithm on GPUs reveals that adding constraints creates an additional bottleneck in the update operation for many real-world sparse tensors. While executing the update operation on the GPU brings significant speedup over its CPU counterpart, it remains a significant bottleneck. To further accelerate the update operation, we propose cuADMM, a new update algorithm that leverages algorithmic and code optimization strategies to minimize both computation and data movement on GPUs. As a result, our framework delivers significantly improved performance compared to prior state-of-the-art. On 10 real-world sparse tensors, our framework achieves geometric mean speedup of 5.1 × (max 41.59 ×) and 7.01 × (max 58.05 ×) on the NIVIDA A100 and H100 GPUs, respectively, over the state-of-the-art SPLATT library running on a 26-core Intel Ice Lake Xeon CPU.

Soh, Yongseok↗

Offloading Calculations to Computational Storage Devices: Spark and HDFS [Slides]

The objective is to evaluate the capabilities of multiple CSDs (provided by NDG Systems) using Hadoop Filesystem and Apache Spark. The independent variables are: number of CSDs, 0, 1, 2, 4, or 6; size of dataset, 1 GB, 5 GB, 10 GB; type of dataset, one large file with all of the data, 10 files, 100 files. The dependent variables are: job time; execution time. the constants are operations on the dataset.

97 MATHEMATICS AND COMPUTING↗

Implementation of High-Order Multireference Coupled-Cluster Methods on Intel Many Integrated Core Architecture

Here we discuss the implementation of multireference coupled-cluster formalism with singles, doubles, and noniterative triples (MRCCSD(T)), which is capable of taking advantage of the processing power of the Intel Xeon Phi coprocessor. We discuss the integration of two levels of parallelism underlying the MRCCSD(T) implementation with computational kernels designed to offload the computationally intensive parts of the MRCCSD(T) formalism to Intel Xeon Phi coprocessors. Special attention is given to the enhancement of the parallel performance by task reordering that has improved load balancing in the noniterative part of the MRCCSD(T) calculations. We also discuss aspects regarding efficient optimization and vectorization strategies.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

10-100 Gbps Offload NIC for WAN, NLR, and Grid Computing

An extremely fast offload engine system has been developed that operates at 60 Gigabits per second (Gbps), and has scalability to 100 Gbps full-duplex (f-d). This system is based on unique coding and architecture derived from splintered UDP (User Datagram Protocol) offload technology, resulting in unique FPGA (field programmable gate array) intellectual property core and firmware. This innovation improves the networking speed of supercomputer clusters by providing an ultra-fast network protocol processing offload from a CPU (central processing unit) by inserting an offload engine into a host backplane and network connections. This runs on protocol firmware.

Awrach, James↗

A Multi-Objective Approach for Optimizing Edge-Based Resource Allocation Using TOPSIS

Existing approaches for allocating resources on edge environments are inefficient and lack the support of heterogeneous edge devices, which in turn fail to optimize the dependency on cloud infrastructures or datacenters. To this extent, we propose in this paper OpERA, a multi-layered edge-based resource allocation optimization framework that supports heterogeneous and seamless execution of offloadable tasks across edge, fog, and cloud computing layers and architectures. By capturing offloadable task requirements, OpERA is capable of identifying suitable resources within nearby edge or fog layers, thus optimizing the execution process. Throughout the paper, we present results which show the effectiveness of our proposed optimization strategy in terms of reducing costs, minimizing energy consumption, and promoting other residual gains in terms of processing computations, network bandwidth, and task execution time. We also demonstrate that by optimizing resource allocation in computation offloading, it is then possible to increase the likelihood of successful task offloading, particularly for computationally intensive tasks that are becoming integral as part of many IoT applications such robotic surgery, autonomous driving, smart city monitoring device grids, and deep learning tasks. The evaluation of our OpERA optimization algorithm reveals that the TOPSIS MCDM technique effectively identifies optimal compute resources for processing offloadable tasks, with a 96% success rate. Moreover, the results from our experiments with a diverse range of use cases show that our OpERA optimization strategy can effectively reduce energy consumption by up to 88%, and operational costs by 76%, by identifying relevant compute resources.

97 MATHEMATICS AND COMPUTING↗

Accelerating Climate and Weather Simulations through Hybrid Computing

Unconventional multi- and many-core processors (e.g. IBM (R) Cell B.E.(TM) and NVIDIA (R) GPU) have emerged as effective accelerators in trial climate and weather simulations. Yet these climate and weather models typically run on parallel computers with conventional processors (e.g. Intel, AMD, and IBM) using Message Passing Interface. To address challenges involved in efficiently and easily connecting accelerators to parallel computers, we investigated using IBM's Dynamic Application Virtualization (TM) (IBM DAV) software in a prototype hybrid computing system with representative climate and weather model components. The hybrid system comprises two Intel blades and two IBM QS22 Cell B.E. blades, connected with both InfiniBand(R) (IB) and 1-Gigabit Ethernet. The system significantly accelerates a solar radiation model component by offloading compute-intensive calculations to the Cell blades. Systematic tests show that IBM DAV can seamlessly offload compute-intensive calculations from Intel blades to Cell B.E. blades in a scalable, load-balanced manner. However, noticeable communication overhead was observed, mainly due to IP over the IB protocol. Full utilization of IB Sockets Direct Protocol and the lower latency production version of IBM DAV will reduce this overhead.

hybrid computing↗