Engineering Papers⌕ Search

Engineering topics

Kerbyson, Darren J.

Publications and source records attributed to Kerbyson, Darren J..

At least 19 records

Scaling deep learning workloads: NVIDIA DGX-1/Pascal and Intel Knights Landing

Deep Learning (DL) algorithms have become ubiquitous in data analytics. As a result, major computing vendors --- including NVIDIA, Intel, AMD, and IBM --- have architectural road-maps influenced by DL workloads. Furthermore, several vendors have recently advertised new computing products as accelerating large DL workloads. Unfortunately, it is difficult for data scientists to quantify the potential of these different products. This paper provides a performance and power analysis of important DL workloads on two major parallel architectures: NVIDIA DGX-1 (eight Pascal P100 GPUs interconnected with NVLink) and Intel Knights Landing (KNL) CPUs interconnected with Intel Omni-Path or Cray Aries. Our evaluation consists of a cross section of convolutional neural net workloads: CifarNet, AlexNet, GoogLeNet, and ResNet50 topologies using the Cifar10 and ImageNet datasets. The workloads are vendor-optimized for each architecture. Our analysis indicates that although GPUs provide the highest overall performance, the gap can close for some convolutional networks; and the KNL can be competitive in performance/watt. We find that NVLink facilitates scaling efficiency on GPUs. However, its importance is heavily dependent on neural network architecture. Furthermore, for weak-scaling --- sometimes encouraged by restricted GPU memory --- NVLink is less important.

Gawande, Nitin A.↗

Generating Performance Models for Irregular Applications

Many applications have irregular behavior --- non-uniform input data, input-dependent solvers, irregular memory accesses, unbiased branches --- that cannot be captured using today's automated performance modeling techniques. We describe new hierarchical critical path analyses for the \Palm model generation tool. To create a model's structure, we capture tasks along representative MPI critical paths. We create a histogram of critical tasks with parameterized task arguments and instance counts. To model each task, we identify hot instruction-level sub-paths and model each sub-path based on data flow, instruction scheduling, and data locality. We describe application models that generate accurate predictions for strong scaling when varying CPU speed, cache speed, memory speed, and architecture. We present results for the Sweep3D neutron transport benchmark; Page Rank on multiple graphs; Support Vector Machine with pruning; and PFLOTRAN's reactive flow/transport solver with domain-induced load imbalance.

analytical performance models↗

Data Provenance Hybridization Supporting Extreme-Scale Scientific WorkflowApplications

As high performance computing (HPC) infrastructures continue to grow in capability and complexity, so do the applications that they serve. HPC and distributed-area computing (DAC) (e.g. grid and cloud) users are looking increasingly toward workflow solutions to orchestrate their complex application coupling, pre- and post-processing needs To gain insight and a more quantitative understanding of a workflow’s performance our method includes not only the capture of traditional provenance information, but also the capture and integration of system environment metrics helping to give context and explanation for a workflow’s execution. In this paper, we describe IPPD’s provenance management solution (ProvEn) and its hybrid data store combining both of these data provenance perspectives.

provenance↗

Combating the Reliability Challenge of GPU Register File at Low Supply Voltage

Supply voltage reduction is an effective approach to significantly reduce GPU energy consumption. As the largest on-chip storage structure, the GPU register file becomes the reliability hotspot that prevents further supply voltage reduction below the safe limit (Vmin) due to process variation effects. This work addresses the reliability challenge of the GPU register file at low supply voltages, which is an essential first step for aggressive supply voltage reduction of the entire GPU chip. We propose GR-Guard, an architectural solution that leverages long register dead time to enable reliable operations from unreliable register file at low voltages.

Architecture-Compiler co-design↗

Assessing Advanced Technology in CENATE

PNNL's Center for Advanced Technology Evaluation (CENATE) is a new U.S. Department of Energy center whose mission is to assess and facilitate access to emerging computing technology. CENATE is assessing a range of advanced technologies, from evolutionary to disruptive. Technologies of interest include the processor socket (homogeneous and accelerated systems), memories (dynamic, static, memory cubes), motherboards, networks (network interface cards and switches), and input/output and storage devices. CENATE is developing a multi-perspective evaluation process based on integrating advanced system instrumentation, performance measurements, and modeling and simulation. We show evaluations of two emerging network technologies: silicon photonics interconnects and the Data Vortex network. CENATE's evaluation also addresses the question of which machine is best for a given workload under certain constraints. We show a performance-power tradeoff analysis of a well-known machine learning application on two systems.

Tallent, Nathan R.↗

New-Sum: A Novel Online ABFT Scheme For General Iterative Methods

Emerging high-performance computing platforms, with large component counts and lower power margins, are anticipated to be more susceptible to soft errors in both logic circuits and memory subsystems. We present an online algorithm-based fault tolerance (ABFT) approach to efficiently detect and recover soft errors for general iterative methods. We design a novel checksum-based encoding scheme for matrix-vector multiplication that is resilient to both arithmetic and memory errors. Our design decouples the checksum updating process from the actual computation, and allows adaptive checksum overhead control. Building on this new encoding mechanism, we propose two online ABFT designs that can effectively recover from errors when combined with a checkpoint/rollback scheme.

Instantaneous Footprint↗

Algorithm and Architecture Independent Benchmarking with SEAK

Many applications of high performance embedded computing are limited by performance or power bottlenecks. We have designed the Suite for Embedded Applications & Kernels (SEAK), a new benchmark suite, (a) to capture these bottlenecks in a way that encourages creative solutions; and (b) to facilitate rigorous, objective, end-user evaluation for their solutions. To avoid biasing solutions toward existing algorithms, SEAK benchmarks use a mission-centric (abstracted from a particular algorithm) and goal-oriented (functional) specification. To encourage solutions that are any combination of software or hardware, we use an end-user black-box evaluation that can capture tradeoffs between performance, power, accuracy, size, and weight. The tradeoffs are especially informative for procurement decisions. We call our benchmarks future proof because each mission-centric interface and evaluation remains useful despite shifting algorithmic preferences. It is challenging to create both concise and precise goal-oriented specifications for mission-centric problems. This paper describes the SEAK benchmark suite and presents an evaluation of sample solutions that highlights power and performance tradeoffs.

algorithms and kernels↗

Fault Modeling of Extreme Scale Applications Using Machine Learning

Faults are commonplace in large scale systems. These systems experience a variety of faults such as transient, permanent and intermittent. Multi-bit faults are typically not corrected by the hardware resulting in an error. Here, this paper attempts to answer an important question: Given a multi-bit fault in main memory, will it result in an application error — and hence a recovery algorithm should be invoked — or can it be safely ignored? We propose an application fault modeling methodology to answer this question. Given a fault signature (a set of attributes comprising of system and application state), we use machine learning to create a model which predicts whether a multibit permanent/transient main memory fault will likely result in error. We present the design elements such as the fault injection methodology for covering important data structures, the application and system attributes which should be used for learning the model, the supervised learning algorithms (and potentially ensembles), and important metrics. Lastly, we use three applications — NWChem, LULESH and SVM — as examples for demonstrating the effectiveness of the proposed fault modeling methodology.

97 MATHEMATICS AND COMPUTING↗

Towards an Application-Specific Thermal Energy Model of Current Processors

Recent developments of high-end processors recognize temperature monitoring and tuning as one of the main challenges towards achieving higher performance given the growing power and temperature constraints. To address this challenge, one needs both suitable thermal energy abstraction and corresponding instrumentation. Our model is based on application-specific parameters such as power consumption, execution time, and asymptotic temperature as well as hardware-specific parameters such as half time for thermal rise or fall. As observed with our out-of-band instrumentation and monitoring infrastructure, the temperature changes follow a relatively slow capacitor-style charge-discharge process. Therefore, we use the lumped thermal model that initiates an exponential process whenever there is a change in processor’s power consumption. Initial experiments with two codes – Firestarter and Nekbone – validate our application-specific thermal model and demonstrate its use for analyzing and potentially improving the balance between temperature, power, and performance.

Getov, Vladimir S.↗

A Case for Application Oblivious Energy-Efficient MPI Runtime

Power has become the major impediment in designing large scale high-end systems. Message Passing Interface (MPI) is the {\em de facto} communication interface used as the back-end for designing applications, programming models and runtime for these systems. Slack --- the time spent by an MPI process in a single MPI call --- provides a potential for energy and power savings, if an appropriate power reduction technique such as core-idling/Dynamic Voltage and Frequency Scaling (DVFS) can be applied without perturbing application's execution time. Existing techniques that exploit slack for power savings assume that application behavior repeats across iterations/executions. However, an increasing use of adaptive, data-dependent workloads combined with system factors (OS noise, congestion) makes this assumption invalid. This paper proposes and implements Energy Aware MPI (EAM) --- an application-oblivious energy-efficient MPI runtime. EAM uses a combination of communication models of common MPI primitives (point-to-point, collective, progress, blocking/non-blocking) and an online observation of slack for maximizing energy efficiency. Each power lever incurs time overhead, which must be amortized over slack to minimize degradation. When predicted communication time exceeds a lever overhead, the lever is used {\em as soon as possible} --- to maximize energy efficiency. When mis-prediction occurs, the lever(s) are used automatically at specific intervals for amortization. We implement EAM using MVAPICH2 and evaluate it on ten applications using up to 4096 processes. Our performance evaluation on an InfiniBand cluster indicates that EAM can reduce energy consumption by 5--41\% in comparison to the default approach, with negligible (less than 4\% in all cases) performance loss.

Venkatesh, Akshay↗

Fast and Accurate Support Vector Machines on Large Scale Systems

Support Vector Machines (SVM) is a supervised Machine Learning and Data Mining (MLDM) algorithm, which has become ubiquitous largely due to its high accuracy and obliviousness to dimensionality. The objective of SVM is to find an optimal boundary --- also known as hyperplane --- which separates the samples (examples in a dataset) of different classes by a maximum margin. Usually, very few samples contribute to the definition of the boundary. However, existing parallel algorithms use the entire dataset for finding the boundary, which is sub-optimal for performance reasons. In this paper, we propose a novel distributed memory algorithm to eliminate the samples which do not contribute to the boundary definition in SVM. We propose several heuristics, which range from early (aggressive) to late (conservative) elimination of the samples, such that the overall time for generating the boundary is reduced considerably. In a few cases, a sample may be eliminated (shrunk) pre-emptively --- potentially resulting in an incorrect boundary. We propose a scalable approach to synchronize the necessary data structures such that the proposed algorithm maintains its accuracy. We consider the necessary trade-offs of single/multiple synchronization using in-depth time-space complexity analysis. We implement the proposed algorithm using MPI and compare it with libsvm--- de facto sequential SVM software --- which we enhance with OpenMP for multi-core/many-core parallelism. Our proposed approach shows excellent efficiency using up to 4096 processes on several large datasets such as UCI HIGGS Boson dataset and Offending URL dataset.

Vishnu, Abhinav↗

Power and Performance Trade-offs for Space Time Adaptive Processing

Computational efficiency – performance relative to power or energy – is one of the most important concerns when designing RADAR processing systems. This paper analyzes power and performance trade-offs for a typical Space Time Adaptive Processing (STAP) application. We study STAP implementations for CUDA and OpenMP on two computationally efficient architectures, Intel Haswell Core I7-4770TE and NVIDIA Kayla with a GK208 GPU. We analyze the power and performance of STAP’s computationally intensive kernels across the two hardware testbeds. We also show the impact and trade-offs of GPU optimization techniques. We show that data parallelism can be exploited for efficient implementation on the Haswell CPU architecture. The GPU architecture is able to process large size data sets without increase in power requirement. The use of shared memory has a significant impact on the power requirement for the GPU. A balance between the use of shared memory and main memory access leads to an improved performance in a typical STAP application.

Gawande, Nitin A.↗

Investigating the Interplay between Energy Efficiency and Resilience in High Performance Computing

Energy efficiency and resilience are two crucial challenges for HPC systems to reach exascale. While energy efficiency and resilience issues have been extensively studied individually, little has been done to understand the interplay between energy efficiency and resilience for HPC systems. Decreasing the supply voltage associated with a given operating frequency for processors and other CMOS-based components can significantly reduce power consumption. However, this often raises system failure rates and consequently increases application execution time. In this work, we present an energy saving undervolting approach that leverages the mainstream resilience techniques to tolerate the increased failures caused by undervolting.

failures↗

Diagnosing the Causes and Severity of One-sided Message Contention

Two trends suggest network contention for one-sided messages is poised to become a performance problem that concerns application developers: an increased interest in one-sided programming models and a rising ratio of hardware threads to network injection bandwidth. Unfortunately, it is difficult to reason about network contention and one-sided messages because one-sided tasks can either decrease or increase contention. We present effective and portable techniques for diagnosing the causes and severity of one-sided message contention. To detect that a message is affected by contention, we maintain statistics representing instantaneous (non-local) network resource demand. Using lightweight measurement and modeling, we identify the portion of a message's latency that is due to contention and whether contention occurs at the initiator or target. We attribute these metrics to program statements in their full static and dynamic context. We characterize contention for an important computational chemistry benchmark on InfiniBand, Cray Aries, and IBM Blue Gene/Q interconnects. We pinpoint the sources of contention, estimate their severity, and show that when message delivery time deviates from an ideal model, there are other messages contending for the same network links. With a small change to the benchmark, we reduce contention up to 50% and improve total runtime as much as 20%.

Network Contention, Performance Analysis, Dynamic ↗

Online Monitoring System for Performance Fault Detection

To achieve the exaFLOPS performance within a contained power budget, next generation supercomputers will feature hundreds of millions of components operating at low- and near-threshold voltage. As the probability that at least one of these components fails during the execution of an application approaches certainty, it seems unrealistic to expect that any run of a scientific application will not experience some performance faults. We believe that there is need of a new generation of light-weight performance and debugging tools that can be used online even during production runs of parallel applications and that can identify performance anomalies during the application execution. In this work we propose the design and implementation of such a monitoring system.

97 MATHEMATICS AND COMPUTING↗

On the Suitability of MPI as a PGAS Runtime

Partitioned Global Address Space (PGAS) models are emerging as a popular alternative to MPI models for designing scalable applications. At the same time, MPI remains a ubiquitous communication subsystem due to its standardization, high performance, and availability on leading platforms. In this paper, we explore the suitability of using MPI as a scalable PGAS communication subsystem. We focus on the Remote Memory Access (RMA) communication in PGAS models which typically includes {\em get, put,} and {\em atomic memory operations}. We perform an in-depth exploration of design alternatives based on MPI. These alternatives include using a semantically-matching interface such as MPI-RMA, as well as not-so-intuitive interfaces such as MPI two-sided with a combination of multi-threading and dynamic process management. With an in-depth exploration of these alternatives and their shortcomings, we propose a novel design which is facilitated by the data-centric view in PGAS models. This design leverages a combination of highly tuned MPI two-sided semantics and an automatic, user-transparent split of MPI communicators to provide asynchronous progress. We implement the asynchronous progress ranks approach and other approaches within the Communication Runtime for Exascale which is a communication subsystem for Global Arrays. Our performance evaluation spans pure communication benchmarks, graph community detection and sparse matrix-vector multiplication kernels, and a computational chemistry application. The utility of our proposed PR-based approach is demonstrated by a 2.17x speed-up on 1008 processors over the other MPI-based designs.

Computational modeling↗

Evaluating Performance and Power Efficiency of Scientific Applications on Multi-threaded Systems

The power and energy walls are changing the way users utilize supercomputers: Time-to-completion is not the only important goal but other metrics, such as the energy required to solve a problem or the power efficiency, are becoming as important as performance. In this paper we analyze the performance, power efficiency and energy consumption of scientific applications programmed in MPI, OpenMP and MPI+OpenMP on two different architectures that have take different approach to limit power consumption, IBM POWER7+ and AMD Interlagos.

Gioiosa, Roberto↗