Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Stacked memory”

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 163 records · Page 9

The Persistent Challenge of Data Locality in the Post-Exascale Era

The era of exascale computing, exemplified by systems like Frontier achieving exaflop-level performance, marks a milestone. However, the quest for sheer compute power leads to strong imbalance in system design. Hence, scaling advancements in memory, network bandwidth, and storage are also necessary and pose challenges, with a crucial need to address data locality issues. This article underscores the fundamental importance of data locality as a key abstraction for optimizing application performance. Despite notable software solutions, the growing complexity of parallelism and memory hierarchy demands performance-portable data locality solutions across diverse computing platforms. Additionally, the article revisits data locality aspects, covering hardware considerations, application perspectives, software stack abstractions, and tool support. It concludes with insights into data locality challenges and opportunities, emphasizing the ongoing significance of collaborative research for progress in this critical issue.

Unat, Didem [Koc University, Istanbul (Turkey)] (O↗

A phase-field model for non-isothermal phase transformation and plasticity in polycrystalline yttria-stabilized tetragonal zirconia

Here we propose an elastoplastic phase-field (PF) model to investigate the mechanics of tetragonal-to-monoclinic phase transformation (TMPT) and elastoplastic deformation of polycrystalline yttria-stabilized tetragonal zirconia (YSTZ). A Landau polynomial with non-vanishing chemical energy at the equilibrium temperature is introduced to account for the actual formation energies of the phases. The effects of different grain orientations, latent heat, and temperature on TMPT and deformation mechanisms are considered. The suppressive transformation effects of the grain boundaries (GBs) is modeled using an inhomogeneous kinetic coefficient in the bulk and GBs. The simulation results for single crystals demonstrate the capability of the model to reproduce the orientation-dependent compressive deformation of YSTZ similar to atomistic simulations and micropillar experiments. The single crystal with [100] crystallographic orientation along the loading direction (SC[100]) displays both TMPT and plasticity, SC[101] experiences only phase transformation, while SC[001] undergoes only plastic yielding. The TMPT induced by compressive loading exhibits shape memory effect (SME) below the equilibrium transformation temperature and pseudoelasticity above it, while the critical transformation stress increases with increasing loading temperature. The irrecoverable plastic strain is found to trap a part of the monoclinic phase, which prevents a complete reverse transformation. The polycrystalline cases also display SME and PE at low and high temperatures, respectively. Due to the orientation differences between grains and the stress concentrations at geometric nonlinearities, plastic deformation occurs in polycrystalline YSTZ for an applied load less than the yield stress. The results suggest a possible limitation of plasticity and an improvement of the shape recovery of YSTZ if one can control the orientation of the grains and/or increase the density of stacking faults at the GBs during material processing.

36 MATERIALS SCIENCE↗

Perpendicular spin transfer torque magnetic random access memories with high spin torque efficiency and thermal stability for embedded applications (invited)

Magnetic random access memories based on the spin transfer torque phenomenon (STT-MRAMs) have become one of the leading candidates for next generation memory applications. Among the many attractive features of this technology are its potential for high speed and endurance, read signal margin, low power consumption, scalability, and non-volatility. In this paper, we discuss our recent results on perpendicular STT-MRAM stack designs that show STT efficiency higher than 5 k{sub B}T/μA, energy barriers higher than 100 k{sub B}T at room temperature for sub-40 nm diameter devices, and tunnel magnetoresistance higher than 150%. We use both single device data and results from 8 Mb array to demonstrate data retention sufficient for automotive applications. Moreover, we also demonstrate for the first time thermal stability up to 400 °C exceeding the requirement of Si CMOS back-end processing, thus opening the realm of non-volatile embedded memory to STT-MRAM technology.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Magnetoresistance of galfenol-based magnetic tunnel junction

The manipulation of ferromagnetic layer magnetization via electrical pulse is driving an intense research due to the important applications that this result will have on memory devices and sensors. In this study we realized a magnetotunnel junction in which one layer is made of Galfenol (Fe{sub 1-x}Ga{sub x}) which possesses one of the highest magnetostrictive coefficient known. The multilayer stack has been grown by molecular beam epitaxy and e-beam evaporation. Optical lithography and physical etching have been combined to obtain 20x20 micron sized pillars. The obtained structures show tunneling conductivity across the junction and a tunnel magnetoresistance (TMR) effect of up to 11.5% in amplitude.

36 MATERIALS SCIENCE↗

Analyzing the Energy and Power Consumption of Remote Memory Accesses in the OpenSHMEM Model

PGAS models like OpenSHMEM provide interfaces to explicitly initiate one-sided remote memory accesses among processes. In addition, the model also provides synchronizing barriers to ensure a consistent view of the distributed memory at different phases of an application. The incorrect use of such interfaces affects the scalability achievable while using a parallel programming model. This study aims at understanding the effects of these constructs on the energy and power consumption behavior of OpenSHMEM applications. Our experiments show that cost incurred in terms of the total energy and power consumed depends on multiple factors across the software and hardware stack. We conclude that there is a significant impact on the power consumed by the CPU and DRAM due to multiple factors including the design of the data transfer patterns within an application, the design of the communication protocols within a middleware, the architectural constraints laid by the interconnect solutions, and also the levels of memory hierarchy within a compute node. This work motivates treating energy and power consumption as important factors while designing compute solutions for current and future distributed systems.

Jana, Siddhartha↗

Analyzing the Energy and Power Consumption of Remote Memory Accesses in the OpenSHMEM Model

PGAS models like OpenSHMEM provide interfaces to explicitly initiate one-sided remote memory accesses among processes. In addition, the model also provides synchronizing barriers to ensure a consistent view of the distributed memory at different phases of an application. The incorrect use of such interfaces affects the scalability achievable while using a parallel programming model. This study aims at understanding the effects of these constructs on the energy and power consumption behavior of OpenSHMEM applications. Our experiments show that cost incurred in terms of the total energy and power consumed depends on multiple factors across the software and hardware stack. We conclude that there is a significant impact on the power consumed by the CPU and DRAM due to multiple factors including the design of the data transfer patterns within an application, the design of the communication protocols within a middleware, the architectural constraints laid by the interconnect solutions, and also the levels of memory hierarchy within a compute node. This work motivates treating energy and power consumption as important factors while designing compute solutions for current and future distributed systems.

Jana, Siddhartha↗

Access Patterns and Performance Behaviors of Multi-layer Supercomputer I/O Subsystems under Production Load

Scientific computing workloads at HPC facilities have been shifting from traditional numerical simulations to AI/ML applications for training and inference while processing and producing ever-increasing amounts of scientific data. To address the growing need for increased storage capacity, lower access latency, and higher bandwidth, emerging technologies such as non-volatile memory are integrated into supercomputer I/O subsystems. With these emerging trends, we need a better understanding of the multilayer supercomputer I/O systems and ways to use these subsystems efficiently. In this work, we study the I/O access patterns and performance characteristics of two representative supercomputer I/O subsystems. Through an extensive analysis of year-long I/O logs on each system, we report new observations in I/O reads and writes, unbalanced use of storage system layers, and new trends in user behaviors at the HPC I/O middleware stack.

Bez, JL↗

Software Requirements for ATDM On-Node Resource Management

This report outlines the software requirements for on-node resource management in the Advanced Simulation and Computing (ASC) Advanced Technology Development and Mitigation (ATDM) project at Sandia National Laboratories (SNL). The need for on-node resource management has arisen from the componentization of the software stack. Componentization aids in managing complexity and making software more composable and reusable. However, components must compete for limited on-node resources for execution (e.g., cores and hardware threads) and memory. The requirements documented in this report support an effort to manage this contention, avoiding oversubscription of resources and enabling their efficient deployment for application execution.

97 MATHEMATICS AND COMPUTING↗

Second-Order Josephson Effect in Excitonic Insulators

We show that in electron-hole bilayers with excitonic orders arising from conduction and valence bands formed by atomic orbitals that have different parities, nonzero interlayer tunneling leads to a second-order Josephson effect. This means the interlayer electrical current is related to the phase of the excitonic order parameter as J = J c sin 2θ instead of J = J c sin θ and that the system has two degenerate ground states at θ = 0, π that can be switched by an interlayer voltage pulse. When generalized to a three dimensional stack of alternating electron-hole planes or a two dimensional stack of chains, the ac Josephson effect implies that electric field pulses perpendicular to the layers and chains can steer the order parameter phase between the two degenerate ground states, making these devices ultrafast memories. Here, the order parameter steering also applies to the excitonic insulator candidate Ta 2 NiSe 5 .

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Quantitative Performance Assessment of Proxy Apps and Parents (ECP Proxy App Project Milestone ADCD-504-9)

This report presents highlights of these efforts. Section 2 describes work that has been done to compare the performance of proxy applications on AMD MI60 vs. Nvidia V100 GPUs. So far only a small set of ECP proxies are running on AMD GPUs, but we will continue to expand this analysis as additional proxies become available. We find that although the MI60 and V100 have nearly the same measured memory bandwidth, memory bound proxy app kernels perform 20-30% worse on the MI60. Further work is needed to refine these comparisons to determine whether the root cause is due to differences in the hardware, software stack, platform specific optimization, or some combination of the three. Section 3 describes our continuing effort to find methods to accurately assess the similarity of proxies and parents. We have recently seen very encouraging results using a cosine similarity metric. This technique uses the angle between two vectors of hardware performance counters to characterize the similarity (or difference) between two applications or proxies. We show not only that several widely used proxies are highly similar to their parents, but also that they differ from non-related codes. We also show that cosine similarity can be used to identify gaps and redundancies in suites and even to gain insight into the effects of architectural differences between platforms. Our work on assessing the Exascale toolchain is ongoing. Our successes with performance measurement tools are evident from the data provided in this report. However, our assessments across the broader tool chain are still too incomplete to provide a meaningful report at this time. We will continue to assess tools and work with vendors and third party developers as issues are identified.

97 MATHEMATICS AND COMPUTING↗

On the error probability of general tree and trellis codes with applications to sequential decoding

An upper bound on the average error probability for maximum-likelihood decoding of the ensemble of random binary tree codes is derived and shown to be independent of the length of the tree. An upper bound on the average error probability for maximum-likelihood decoding of the ensemble of random L-branch binary trellis codes of rate R = 1/n is derived which separates the effects of the tail length T and the memory length M of the code. It is shown that the bound is independent of the length L of the information sequence. This implication is investigated by computer simulations of sequential decoding utilizing the stack algorithm. These simulations confirm the implication and further suggest an empirical formula for the true undetected decoding error probability with sequential decoding.

Johannesson, R.↗

Thread Migration in the Presence of Pointers

Dynamic migration of lightweight threads supports both data locality and load balancing. However, migrating threads that contain pointers referencing data in both the stack and heap remains an open problem. In this paper we describe a technique by which threads with pointers referencing both stack and non-shared heap data can be migrated such that the pointers remain valid after migration. As a result, threads containing pointers can now be migrated between processors in a homogeneous distributed memory environment.

Cronk, David↗

Phase-change heterostructure enables ultralow noise and drift for memory operation

Artificial intelligence and other data-intensive applications have escalated the demand for data storage and processing. New computing devices, such as phase-change random access memory (PCRAM)–based neuro-inspired devices, are promising options for breaking the von Neumann barrier by unifying storage with computing in memory cells. However, current PCRAM devices have considerable noise and drift in electrical resistance that erodes the precision and consistency of these devices. We designed a phase-change heterostructure (PCH) that consists of alternately stacked phase-change and confinement nanolayers to suppress the noise and drift, allowing reliable iterative RESET and cumulative SET operations for high-performance neuro-inspired computing. Our PCH architecture is amenable to industrial production as an intrinsic materials solution, without complex manufacturing procedure or much increased fabrication cost.

Science & Technology - Other Topics↗

SpaceCube v2.0 Space Flight Hybrid Reconfigurable Data Processing System

This paper details the design architecture, design methodology, and the advantages of the SpaceCube v2.0 high performance data processing system for space applications. The purpose in building the SpaceCube v2.0 system is to create a superior high performance, reconfigurable, hybrid data processing system that can be used in a multitude of applications including those that require a radiation hardened and reliable solution. The SpaceCube v2.0 system leverages seven years of board design, avionics systems design, and space flight application experiences. This paper shows how SpaceCube v2.0 solves the increasing computing demands of space data processing applications that cannot be attained with a standalone processor approach.The main objective during the design stage is to find a good system balance between power, size, reliability, cost, and data processing capability. These design variables directly impact each other, and it is important to understand how to achieve a suitable balance. This paper will detail how these critical design factors were managed including the construction of an Engineering Model for an experiment on the International Space Station to test out design concepts. We will describe the designs for the processor card, power card, backplane, and a mission unique interface card. The mechanical design for the box will also be detailed since it is critical in meeting the stringent thermal and structural requirements imposed by the processing system. In addition, the mechanical design uses advanced thermal conduction techniques to solve the internal thermal challenges.The SpaceCube v2.0 processing system is based on an extended version of the 3U cPCI standard form factor where each card is 190mm x 100mm in size The typical power draw of the processor card is 8 to 10W and scales with application complexity. The SpaceCube v2.0 data processing card features two Xilinx Virtex-5 QV Field Programmable Gate Arrays (FPGA), eight memory modules, a monitor FPGA with analog monitoring, Ethernet, configurable interconnect to the Xilinx FPGAs including gigabit transceivers, and the necessary voltage regulation. The processor board uses a back-to-back design methodology for common parts that maximizes the board real estate available. This paper will show how to meet the IPC 6012B Class 3A standard with a 22-layer board that has two column grid array devices with 1.0mm pitch. All layout trades such as stack-up options, via selection, and FPGA signal breakout will be discussed with feature size results. The overall board design process will be discussed including parts selection, circuit design, proper signal termination, layout placement and route planning, signal integrity design and verification, and power integrity results. The radiation mitigation techniques will also be detailed including configuration scrubbing options, Xilinx circuit mitigation and FPGA functional monitoring, and memory protection.Finally, this paper will describe how this system is being used to solve the extreme challenges of a robotic satellite servicing mission where typical space-rated processors are not sufficient enough to meet the intensive data processing requirements. The SpaceCube v2.0 is the main payload control computer and is required to control critical subsystems such as autonomous rendezvous and docking using a suite of vision sensors and object avoidance when controlling two robotic arms.

Xilinx FPGA↗

In-Memory Graph Databases for Web-Scale Data

RDF databases have emerged as one of the most relevant way for organizing, integrating, and managing expo- nentially growing, often heterogeneous, and not rigidly structured data for a variety of scientific and commercial fields. In this paper we discuss the solutions integrated in GEMS (Graph database Engine for Multithreaded Systems), a software framework for implementing RDF databases on commodity, distributed-memory high-performance clusters. Unlike the majority of current RDF databases, GEMS has been designed from the ground up to primarily employ graph-based methods. This is reflected in all the layers of its stack. The GEMS framework is composed of: a SPARQL-to-C++ compiler, a library of data structures and related methods to access and modify them, and a custom runtime providing lightweight software multithreading, network messages aggregation and a partitioned global address space. We provide an overview of the framework, detailing its component and how they have been closely designed and customized to address issues of graph methods applied to large-scale datasets on clusters. We discuss in details the principles that enable automatic translation of the queries (expressed in SPARQL, the query language of choice for RDF databases) to graph methods, and identify differences with respect to other RDF databases.

GEMS↗

WfBench: Automated Generation of Scientific Workflow Benchmarks

The prevalence of scientific workflows with high computational demands calls for their execution on various distributed computing platforms, including large-scale leadership-class high-performance computing (HPC) clusters. To handle the deployment, monitoring, and optimization of workflow executions, many workflow systems have been developed over the past decade. There is a need for workflow benchmarks that can be used to evaluate the performance of workflow systems on current and future software stacks and hardware platforms.We present a generator of realistic workflow benchmark specifications that can be translated into benchmark code to be executed with current workflow systems. Our approach generates workflow tasks with arbitrary performance characteristics (CPU, memory, and I/O usage) and with realistic task dependency structures based on those seen in production workflows. We present experimental results that show that our approach generates benchmarks that are representative of production workflows, and conduct a case study to demonstrate the use and usefulness of our generated benchmarks to evaluate the performance of workflow systems under different configuration scenarios.

Coleman, Taina↗

Two-terminal electronic charge resistance switching device

A two-terminal memory device and methods for its use are provided. In the device, a bottom electrode is electrically continuous with a first operating terminal, and a control gate electrode is electrically continuous with a second operating terminal. A stack of insulator layers comprising a hopping conduction layer and a tunnel layer is contactingly interposed between the bottom electrode and the control gate electrode. The tunnel layer is thinner than the hopping conduction layer, and it has a wider bandgap than the hopping conduction layer. The hopping conduction layer consists of a material that supports electron hopping transport.

Marinella, Matthew↗

Analyzing inference workloads for spatiotemporal modeling

Ensuring power grid resiliency, forecasting climate conditions, and optimization of transportation infrastructure are some of the many application areas where data is collected in both space and time. Spatiotemporal modeling is about modeling those patterns for forecasting future trends and carrying out critical decision-making by leveraging machine learning/deep learning. Once trained offline, field deployment of trained models for near real-time inference could be challenging because performance can vary significantly depending on the environment, available compute resources and tolerance to ambiguity in results. Users deploying spatiotemporal models for solving complex problems can benefit from analytical studies considering a plethora of system adaptations to understand the associated performance-quality trade-offs. To facilitate the co-design of next-generation hardware architectures for field deployment of trained models, it is critical to characterize the workloads of these deep learning (DL) applications during inference and assess their computational patterns at different levels of the execution stack. In this paper, we develop several variants of deep learning applications that use spatiotemporal data from dynamical systems. We study the associated computational patterns for inference workloads at different levels, considering relevant models (Long short-term Memory, Convolutional Neural Network and Spatio-Temporal Graph Convolution Network), DL frameworks (Tensorflow and PyTorch), precision (FP16, FP32, AMP, INT16 and INT8), inference runtime (ONNX and AI Template), post-training quantization (TensorRT) and platforms (Nvidia DGX A100 and Sambanova SN10 RDU). Overall, our findings indicate that although there is potential in mixed-precision models and post-training quantization for spatiotemporal modeling, extracting efficiency from contemporary GPU systems might be challenging. Instead, co-designing custom accelerators by leveraging optimized High Level Synthesis frameworks (such as SODA High-Level Synthesizer for customized FPGA/ASIC targets) can make workload-specific adjustments to enhance the efficiency.

97 MATHEMATICS AND COMPUTING↗