Search NASA⌕ Search

SEARCH · Search NASA

Results for “synchronization overhead”

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 91 records · Page 5

Three practical workflow schedulers for easy maximum parallelism

Runtime scheduling and workflow systems are an increasingly popular algorithmic component in HPC because they allow full system utilization with relaxed synchronization requirements. There are so many special-purpose tools for task scheduling, one might wonder why more are needed. Use cases seen on the Summit supercomputer needed better integration with MPI and greater flexibility in job launch configurations. Preparation, execution, and analysis of computational chemistry simulations at the scale of tens of thousands of processors revealed three distinct workflow patterns. A separate job scheduler was implemented for each one using extremely simple and robust designs: file-based, task-list based, and bulk-synchronous. Comparing to existing methods shows unique benefits of this work, including simplicity of design, suitability for HPC centers, short startup time, and well-understood per-task overhead. All three new tools have been shown to scale to full utilization of Summit, and have been made publicly available with tests and documentation. This work presents a complete characterization of the minimum effective task granularity for efficient scheduler usage scenarios. Here, these schedulers have the same bottlenecks, and hence similar task granularities as those reported for existing tools following comparable paradigms.

97 MATHEMATICS AND COMPUTING↗

Parallel Breadth-First Search on Distributed Memory Systems

Data-intensive, graph-based computations are pervasive in several scientific applications, and are known to be quite challenging to implement on distributed memory systems. In this work, we explore the design space of parallel algorithms for Breadth-First Search (BFS), a key subroutine in several graph algorithms. We present two highly-tuned parallel approaches for BFS on large parallel systems: a level-synchronous strategy that relies on a simple vertex-based partitioning of the graph, and a two-dimensional sparse matrix-partitioning-based approach that mitigates parallel communication overhead. For both approaches, we also present hybrid versions with intra-node multithreading. Our novel hybrid two-dimensional algorithm reduces communication times by up to a factor of 3.5, relative to a common vertex based approach. Our experimental study identifies execution regimes in which these approaches will be competitive, and we demonstrate extremely high performance on leading distributed-memory parallel systems. For instance, for a 40,000-core parallel execution on Hopper, an AMD Magny-Cours based system, we achieve a BFS performance rate of 17.8 billion edge visits per second on an undirected graph of 4.3 billion vertices and 68.7 billion edges with skewed degree distribution.

97 MATHEMATICS AND COMPUTING↗

A Network Contention Model for the Extreme-scale Simulator

The Extreme-scale Simulator (xSim) is a performance investigation toolkit for high-performance computing (HPC) hardware/software co-design. It permits running a HPC application with millions of concurrent execution threads, while observing its performance in a simulated extreme-scale system. This paper details a newly developed network modeling feature for xSim, eliminating the shortcomings of the existing network modeling capabilities. The approach takes a different path for implementing network contention and bandwidth capacity modeling using a less synchronous and accurate enough model design. With the new network modeling feature, xSim is able to simulate on-chip and on-node networks with reasonable accuracy and overheads.

Engelmann, Christian↗

Multithreaded Model for Dynamic Load Balancing Parallel Adaptive PDE Computations

We present a multithreaded model for the dynamic load-balancing of numerical, adaptive computations required for the solution of Partial Differential Equations (PDE's) on multiprocessors. Multithreading is used as a means of exploring concurrency in the processor level in order to tolerate synchronization costs inherent to traditional (non-threaded) parallel adaptive PDE solvers. Our preliminary analysis for parallel, adaptive PDE solvers indicates that multithreading can be used an a mechanism to mask overheads required for the dynamic balancing of processor workloads with computations required for the actual numerical solution of the PDE's. Also, multithreading can simplify the implementation of dynamic load-balancing algorithms, a task that is very difficult for traditional data parallel adaptive PDE computations. Unfortunately, multithreading does not always simplify program complexity, often makes code re-usability not an easy task, and increases software complexity.

Chrisochoides, Nikos↗

Memristive Mixed-Signal Neuromorphic Systems: Energy-Efficient Learning at the Circuit-Level

Neuromorphic computing is non-von Neumann computer architecture for the post Moore’s law era of computing. Since a main focus of the post Moore’s law era is energy-efficient computing with fewer resources and less area, neuromorphic computing contributes effectively in this research. Here in this paper, we present a memristive neuromorphic system for improved power and area efficiency. Our particular mixed-signal approach implements neural networks with spiking events in a synchronous way. Moreover, the use of nano-scale memristive devices saves both area and power in the system. We also provide device-level considerations that make the system more energy-efficient. The proposed system additionally includes synchronous digital long term plasticity, an online learning methodology that helps the system train the neural networks during the operation phase and improves the efficiency in learning considering the power consumption and area overhead.

77 NANOSCIENCE AND NANOTECHNOLOGY↗

Update transport - A new technique for update synchronization in replicated database systems

A fully distributed approach to update synchronization is presented where each site completely executes every update. This approach has several features-higher resiliency to different kinds of failures, higher parallelism, improved response to user requests, and low communication overhead. A fully distributed algorithm for concurrency control obtained by rehashing a previously published semidistributed algorithm into the fully distributed model of update execution is presented. A performance model of replicated database systems is presented and used to study the performance of the proposed algorithm and its semidistributed version. The results of the performance study reveal that the proposed approach can substantially improve the performance at the cost of moderate input/output overhead.

Singhal, Mukesh↗

Complete exchange on the iPSC-860

The implementation of complete exchange on the circuit switched Intel iPSC-860 hypercube is described. This pattern, also known as all-to-all personalized communication, is the densest requirement that can be imposed on a network. On the iPSC-860, care needs to be taken to avoid edge contention, which can have a disastrous impact on communication time. There are basically two classes of algorithms that achieve contention-free complete exchange. The first contains the classical standard exchange algorithm that is generally useful for small message sizes. The second includes a number of optimal or near-optimal algorithms that are best for large messages. Measurement of communication overhead on the iPSC-860 are given and a notation for analyzing communication link usage is developed. It is shown that for the two classes of algorithms, there is substantial variation in performance with synchronization technique and choice of message protocol. Timings of six implementations are given; each of these is useful over a particular range of message size and cube dimension. Since the complete exchange is a superset of communication patterns, these timings represent upper bounds on the time required by an arbitrary communication requirement. These results indicate that the programmer needs to evaluate several possibilities before finalizing an implementation - a careful choice can lead to very significant savings in time.

Bokhari, Shahid H.↗

Energy–Performance Trade-offs in Privacy-Preserving Federated Learning on SmartNIC-Enabled HPC Systems

Federated learning (FL) is increasingly deployed on accelerator-rich high-performance computing (HPC) systems, yet the system-level energy cost of privacy-aware FL remains poorly understood, particularly across heterogeneous networking and server-placement options. We present a measurement-driven study of energy–performance trade-offs for FL on GH200-class nodes across three deployment configurations: CPU-Ethernet, CPU-InfiniBand (RDMA-capable), and a DPU-hosted FL server over InfiniBand using a BlueField-3 SmartNIC/DPU. Using NVIDIA FLARE (NVFLARE), we align node-level power telemetry with per-round timing extracted from NVFLARE logs to quantify time-to-solution (TTS), energy-to-solution (ETS), energy-delay product (EDP), and synchronization behavior for three transformer models (ALBERT, DistilBERT, BERT), trained with and without differential privacy (DP). We find that interconnect choice is the dominant driver of runtime and energy: host-managed InfiniBand consistently reduces communication overhead versus Ethernet, yielding lower TTS/ETS/EDP. In contrast, in our NVFLARE deployment, placing the FL server on the DPU does not consistently match CPU-InfiniBand performance and can be slower—especially for larger models—highlighting that server placement alone is not sufficient to guarantee end-to-end gains. Finally, under our fixed-round protocol, DP increases per-round cost and runtime variance; ETS increases largely in proportion to TTS because average node power remains relatively stable across configurations.

Kotevska, Olivera [ORNL] (ORCID:0000000316772243)↗

Performance bounds on parallel self-initiating discrete-event

The use is considered of massively parallel architectures to execute discrete-event simulations of what is termed self-initiating models. A logical process in a self-initiating model schedules its own state re-evaluation times, independently of any other logical process, and sends its new state to other logical processes following the re-evaluation. The interest is in the effects of that communication on synchronization. The performance is considered of various synchronization protocols by deriving upper and lower bounds on optimal performance, upper bounds on Time Warp's performance, and lower bounds on the performance of a new conservative protocol. The analysis of Time Warp includes the overhead costs of state-saving and rollback. The analysis points out sufficient conditions for the conservative protocol to outperform Time Warp. The analysis also quantifies the sensitivity of performance to message fan-out, lookahead ability, and the probability distributions underlying the simulation.

Nicol, David M.↗

Exploring Hardware Support For Scaling Irregular Applications on Multi-node Multi-core Architectures

With the recent emergence of large-scale knowledge dis- covery, data mining and social network analysis, irregular applications have gained renewed interest. Classic cache-based high-performance architectures do not provide optimal performances with such kind of workloads, mainly due to the very low spatial and temporal locality of the irregular control and memory access patterns. In this paper, we present a multi-node, multi-core, fine-grained multi-threaded shared-memory system architecture specifically designed for the execution of large-scale irregular applications, and built on top of three pillars, that we believe are fundamental to support these workloads. First, we offer transparent hardware support for Partitioned Global Address Space (PGAS) to provide a large globally-shared address space with no software library overhead. Second, we employ multi-threaded multi-core processing nodes to achieve the necessary latency tolerance required by accessing global memory, which potentially resides in a remote node. Finally, we devise hardware support for inter-thread synchronization on the whole global address space. We first model the performances by using an analytical model that takes into account the main architecture and application characteristics. We describe the hardware design of the proposed cus- tom architectural building blocks that provide support for the above- mentioned three pillars. Finally, we present a limited-scale evaluation of the system on a multi-board FPGA prototype with typical irregular kernels and benchmarks. The experimental evaluation demonstrates the architecture performance scalability for different configurations of the whole system.

FPGA↗

A Synchronization Algorithm and Implementation for High-Speed Block Codes Applications

Block codes have trellis structures and decoders amenable to high speed CMOS VLSI implementation. For a given CMOS technology, these structures enable operating speeds higher than those achievable using convolutional codes for only modest reductions in coding gain. As a result, block codes have tremendous potential for satellite trunk and other future high-speed communication applications. This paper describes a new approach for implementation of the synchronization function for block codes. The approach utilizes the output of the Viterbi decoder and therefore employs the strength of the decoder. Its operation requires no knowledge of the signal-to-noise ratio of the received signal, has a simple implementation, adds no overhead to the transmitted data, and has been shown to be effective in simulation for received SNR greater than 2 dB.

Lin, Shu↗

Fault-tolerance for exascale systems.

Periodic, coordinated, checkpointing to disk is the most prevalent fault tolerance method used in modern large-scale, capability class, high-performance computing (HPC) systems. Previous work has shown that as the system grows in size, the inherent synchronization of coordinated checkpoint/restart (CR) limits application scalability; at large node counts the application spends most of its time checkpointing instead of executing useful work. Furthermore, a single component failure forces an application restart from the last correct checkpoint. Suggested alternatives to coordinated CR include uncoordinated CR with message logging, redundant computation, and RAID-inspired, in-memory distributed checkpointing schemes. Each of these alternatives have differing overheads that are dependent on both the scale and communication characteristics of the application. In this work, using the Structural Simulation Toolkit (SST) simulator, we compare the performance characteristics of each of these resilience methods for a number of HPC application patterns on a number of proposed exascale machines. The result of this work provides valuable guidance on the most efficient resilience methods for exascale systems.

99 GENERAL AND MISCELLANEOUS↗

Daytime Cognitive Performance in Response to Sunlight or Fluorescent Light Controlling for Sleep Duration

Light is the primary synchronizer of the human circadian rhythm and also has acute alerting effects. Our study involves and comparing the alertness, performance and sleep of participants in the NASA Ames Sustainability Base, which uses sunlight as its primary light source, to in a traditional office building which uses overhead florescent lighting and varying exposure to natural light. The purpose of this study is to determine whether the use of natural lighting as a primary light source improves daytime cognitive function and promotes nighttime sleep. Participants from the Sustainability Base will be matched by gender and age to individuals working in other NASA buildings. In a prior study we found no differences in performance between those working in the Sustainability Base and those working in other buildings. Unexpectedly, we found that the average sleep duration among participants in both buildings was short, which likely obscured our ability to detect a difference the effect of light exposure on alertness. Given that such sleep deprivation has negative effects on cognitive performance, in this iteration of the study we are asking the participants to maintain a regular schedule with eight hours in bed each night in order to control for the effect of self-selected sleep restriction. Over the course of one week, we will ask the participants to wear actiwatches continuously, complete a psychomotor vigilance task (PVT) and digit symbol substitution task (DSST) three times per day, and keep daily sleepwork diaries. We hope that this study will provide data to support the idea that natural lighting and green architectural design are optimal to enhance healthy nighttime sleep patterns and daytime cognitive performance.

cognitive performance↗

Juggler: a dependence-aware task-based execution framework for GPUs

Scientific applications with single instruction, multiple data (SIMD) computations show considerable performance improvements when run on today's graphics processing units (GPUs). However, the existence of data dependences across thread blocks may significantly impact the speedup by requiring global synchronization across multiprocessors (SMs) inside the GPU. To efficiently run applications with interblock data dependences, we need fine-granular task-based execution models that will treat SMs inside a GPU as stand-alone parallel processing units. Such a scheme will enable faster execution by utilizing all internal computation elements inside the GPU and eliminating unnecessary waits during device-wide global barriers.In this paper, we propose Juggler, a task-based execution scheme for GPU workloads with data dependences. The Juggler framework takes applications embedding OpenMP 4.5 tasks as input and executes them on the GPU via an efficient in-device runtime, hence eliminating the need for kernel-wide global synchronization. Juggler requires no or little modification to the source code, and once launched, the runtime entirely runs on the GPU without relying on the host through the entire execution. We have evaluated Juggler on an NVIDIA Tesla P100 GPU and obtained up to 31% performance improvement against global barrier based implementation, with minimal runtime overhead.

Belviranli, Mehmet E.↗

Distributed Order Recording Techniques for Efficient Record-and-Replay of Multi-threaded Programs

After all these years and all these other shared memory programming frameworks, OpenMP is still the most popular one. However, its greater levels of non-deterministic execution makes debugging and testing more challenging. The ability to record and deterministically replay the program execution is key to address this challenge. However, scalably replaying OpenMP programs is still an unresolved problem. In this paper, we propose two novel techniques that use Distributed Clock (DC) and Distributed Epoch (DE) recording schemes to eliminate excessive thread synchronization for OpenMP record and replay. Our evaluation on representative HPC applications with ReOMP, which we used to realize DC and DE recording, shows that our approach is 2-5x more efficient than traditional approaches that synchronize on every shared-memory access. Furthermore, we demonstrate that our approach can be easily combined with MPI-level replay tools to replay non-trivial MPI+OpenMP applications. We achieve this by integrating ReOMP into ReMPI, an existing scalable MPI record-and-replay tool, with only a small MPI-scale-independent runtime overhead.

Fu, Xiang↗

Code modernization strategies for short-range non-bonded molecular dynamics simulations

Modern HPC systems are increasingly relying on greater core counts and wider vector registers. Thus, applications need to be adapted to fully utilize these hardware capabilities. One class of applications that can benefit from this increase in parallelism are molecular dynamics simulations. In this paper, we describe our efforts at modernizing the ESPResSo++ simulation package for molecular dynamics by restructuring its particle data layout for efficient memory accesses and applying vectorization techniques to benefit the calculation of short-range non-bonded forces, which results in an overall three times speedup and serves as a baseline for further optimizations. We also implement fine-grained parallelism for multi-core CPUs through HPX, a C++ runtime system which uses lightweight threads and an asynchronous many-task approach to maximize concurrency. Our goal is to evaluate the performance of an HPX-based approach compared to the bulk-synchronous MPI-based implementation. This requires the introduction of an additional layer to the domain decomposition scheme that defines the task granularity. On spatially inhomogeneous systems, which impose a corresponding load-imbalance in traditional MPI-based approaches, we demonstrate that by choosing an optimal task size, the efficient work-stealing mechanisms of HPX can overcome the overhead of communication resulting in an overall 1.4 times speedup compared to the baseline MPI version.

97 MATHEMATICS AND COMPUTING↗

Smart and Intelligent Sensors

John C. Stennis Space Center (SSC) provides rocket engine propulsion testing for NASA's space programs. Since the development of the Space Shuttle, every Space Shuttle Main Engine (SSME) has undergone acceptance testing at SSC before going to Kennedy Space Center (KSC) for integration into the Space Shuttle. The SSME is a large cryogenic rocket engine that uses Liquid Hydrogen (LH2) as the fuel. As NASA moves to the new ARES V launch system, the main engines on the new vehicle, as well as the upper stage engine, are currently base lined to be cryogenic rocket engines that will also use LH2. The main rocket engines for the ARES V will be larger than the SSME, while the upper stage engine will be approximately half that size. As a result, significant quantities of hydrogen will be required during the development, testing, and operation of these rocket engines.Better approaches are needed to simplify sensor integration and help reduce life-cycle costs. 1.Smarter sensors. Sensor integration should be a matter of "plug-and-play" making sensors easier to add to a system. Sensors that implement new standards can help address this problem; for example, IEEE STD 1451.4 defines transducer electronic data sheet (TEDS) templates for commonly used sensors such as bridge elements and thermocouples. When a 1451.4 compliant smart sensor is connected to a system that can read the TEDS memory, all information needed to configure the data acquisition system can be uploaded. This reduces the amount of labor required and helps minimize configuration errors. 2.Intelligent sensors. Data received from a sensor be scaled, linearized; and converted to engineering units. Methods to reduce sensor processing overhead at the application node are needed. Smart sensors using low-cost microprocessors with integral data acquisition and communication support offer the means to add these capabilities. Once a processor is embedded, other features can be added; for example, intelligent sensors can make a health assessment to inform the data acquisition client when sensor performance is suspect. 3.Distributed sample synchronization. Networks of sensors require new ways for synchronizing samples. Standards that address the distributed timing problem (for example, IEEE STD 1588) provide the means to aggregate samples from many distributed smart sensors with sub-microsecond accuracy. 4. Reduction in interconnect. Alternative means are needed to reduce the frequent problems associated with cabling and connectors. Wireless technologies offer the promise of reducing interconnects and simultaneously making it easy to quickly add a sensor to a system.

Lansaw, John↗

Lightweight causal and atomic group multicast

The ISIS toolkit is a distributed programming environment based on support for virtually synchronous process groups and group communication. A suite of protocols is presented to support this model. The approach revolves around a multicast primitive, called CBCAST, which implements a fault-tolerant, causally ordered message delivery. This primitive can be used directly or extended into a totally ordered multicast primitive, called ABCAST. It normally delivers messages immediately upon reception, and imposes a space overhead proportional to the size of the groups to which the sender belongs, usually a small number. It is concluded that process groups and group communication can achieve performance and scaling comparable to that of a raw message transport layer. This finding contradicts the widespread concern that this style of distributed computing may be unacceptably costly.

Birman, Kenneth P.↗