Memphis on an XT5: Pinpointing Memory Performance Problems on Cray Platforms
No abstract prepared.
SEARCH · Search NASA
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.
No abstract prepared.
Abstract Not Provided
Abstract Not Provided
Abstract not provided.
We address memory performance issues observed in Blue Gene Linux and discuss the design and implementation of 'Big Memory' - an alternative, transparent memory space introduced to eliminate the memory performance issues. We evaluate the performance of Big Memory using custom memory benchmarks, NAS Parallel Benchmarks, and the Parallel Ocean Program, at a scale of up to 4,096 nodes. We find that Big Memory successfully resolves the performance issues normally encountered in Blue Gene Linux. For the ocean simulation program, we even find that Linux with Big Memory provides better scalability than does the lightweight compute node kernel designed solely for high-performance applications. Originally intended exclusively for compute node tasks, our new memory subsystem dramatically improves the performance of certain I/O node applications as well. We demonstrate this performance using the central processor of the LOw Frequency ARray radio telescope as an example.
The k-dimensional Weisfeiler-Lehman (k-WL) algorithm—developed as an efficient heuristic for testing if two graphs are isomorphic—is a fundamental kernel for node embedding in the emerging field of graph neural networks. Unfortunately, the k-WL algorithm has exponential storage requirements, limiting the size of graphs that can be handled. This work presents a novel k-WL scheme with a storage requirement orders of magnitude lower while maintaining the same accuracy as the original k-WL algorithm. Due to the reduced storage requirement, our scheme allows for processing much bigger graphs than previously possible on a single compute node. For even bigger graphs, we provide the first distributed-memory implementation. Our k-WL scheme also has significantly reduced communication volume and offers high scalability. Our experimental results demonstrate that our approach is significantly faster and has superior scalability compared to five other implementations employing state-of-the-art techniques.
A method and apparatus of performing a memory operation includes receiving a memory operation request at a first memory controller that is in communication with a second memory controller. The first memory controller forwards the memory operation request to the second memory controller. Upon receipt of the memory operation request, the second memory controller provides first information or second information depending on a condition of a pseudo-bank of the second memory controller and a type of the memory operation request.
This project investigated methods for using recordings of event-related potentials (ERPs) to predict and improve human memory performance. In a series of experiments, the ERPs related to good and poor memory performance were characterized and a computational model was developed in which the ERPs recorded while a person studied words were used to predict whether that person would be a high performer or a low performer on a subsequent memory test. These experiments provide the first demonstration of ERPs used to predict subsequent memory performance. Additional experiments investigated the effects of different types of memory training (learning a memory strategy such as mental imagery or increasing memory capacity through working memory training) on memory performance and the associated brain activity.
Associative memory refers to remembering the association between two items, such as a face and a name. It is a crucial part of daily life, but it is also one of the first aspects of memory performance that is impacted by aging and by Alzheimer’s disease. Evidence suggests that transcranial direct current stimulation (tDCS) can improve memory performance, but few tDCS studies have investigated its impact on associative memory. In addition, no prior study of the effects of tDCS on memory performance has systematically evaluated the impact of tDCS on different types of memory assessments, such as recognition and recall tests. In this study, we measured the effects of tDCS on associative memory performance in healthy adults, using both recognition and recall tests. Participants studied face-name pairs while receiving either active (30 minutes, 2 mA) or sham (30 minutes, 0.1 mA) stimulation with the anode placed at F9 and the cathode placed on the contralateral upper arm. Participants in the active stimulation group performed significantly better on the recall test than participants in the sham group, recalling 50% more names, on average, and making fewer recall errors. However, the two groups did not differ significantly in terms of their performance on the recognition memory test. This investigation provides evidence that stimulation at the time of study improves associative memory encoding, but that this memory benefit is evident only under certain retrieval conditions.
MRAM and DRAM memory technologies identified for possible radiation hardening were investigated for memory-array- and controller-level radiation sensitivity. TID and SEE performance of memory arrays and associated cells is good, but with some questions about viability of hardening of problematic control circuitry. Control-circuitry-related SEE problems are highlighted. TID data are focused on memory arrays only.
The disparity between microprocessor clock frequencies and memory latency is a primary reason why many demanding applications run well below peak achievable performance. Software controlled scratchpad memories, such as the Cell local store, attempt to ameliorate this discrepancy by enabling precise control over memory movement; however, scratchpad technology confronts the programmer and compiler with an unfamiliar and difficult programming model. In this work, we present the Virtual Vector Architecture (ViVA), which combines the memory semantics of vector computers with a software-controlled scratchpad memory in order to provide a more effective and practical approach to latency hiding. ViVA requires minimal changes to the core design and could thus be easily integrated with conventional processor cores. To validate our approach, we implemented ViVA on the Mambo cycle-accurate full system simulator, which was carefully calibrated to match the performance on our underlying PowerPC Apple G5 architecture. Results show that ViVA is able to deliver significant performance benefits over scalar techniques for a variety of memory access patterns as well as two important memory-bound compact kernels, corner turn and sparse matrix-vector multiplication -- achieving 2x-13x improvement compared the scalar version. Overall, our preliminary ViVA exploration points to a promising approach for improving application performance on leading microprocessors with minimal design and complexity costs, in a power efficient manner.
Prior work demonstrates that application of transcranial direct current stimulation (tDCS) improves memory. In this study, we investigated tDCS effects on face-name associative memory using both recall and recognition tests. Participants encoded face-name pairs under either active (1.5 mA) or sham (.1 mA) stimulation applied to the scalp adjacent to the left dorsolateral prefrontal cortex (dlPFC), an area known to support associative memory. Participants’ memory was then tested after study (day one) and then again after a 24-h delay (day two), to assess both immediate and delayed stimulation effects on memory. Results indicated that active relative to sham stimulation led to substantially improved recall (more than 50%) at both day one and day two. Recognition memory performance did not differ between stimulation groups at either time point. These results suggest that stimulation at encoding improves memory performance by enhancing memory for details that enable a rich recollective experience, but that these improvements are evident only under some testing conditions, especially those that rely on recollection. Altogether, stimulation of the dlPFC could have led to recall improvement through enhanced encoding from stimulation or from carryover effects of stimulation that influenced retrieval processes, or both.
In the age of massive parallelism, the focus of performance analysis has switched from the processor and related structures to the memory and I/O resources. Adapting to this new reality, a performance analysis tool has to provide a way to analyze resource usage to pinpoint existing and potential problems in a given application. This paper provides an overview of the Memory Observant Data Analysis (MODA) tool, a memory-centric tool first implemented on the Cray XMT supercomputer. Throughout the paper, MODA's capabilities have been showcased with experiments done on matrix multiply and Graph-500 application codes.
In order to increase the thermal stability of a magnetic random access memory cell, materials with high spin-orbit interaction are often introduced in the storage layer. As a side effect, a strong Dzyaloshinskii-Moriya interaction (DMI) may arise in such systems. Here, we investigate the impact of DMI on the magnetic cell performance, using micromagnetic simulations. We find that DMI strongly promotes non-uniform magnetization states and non-uniform switching modes of the magnetic layer. It appears to be detrimental for both the thermal stability of the cell and its switching current, leading to considerable deterioration of the cell performance even for a moderate DMI amplitude.
An unprecedented four-dimensional (4D) printing process allowing high-performance and shape memory thermoset to be printed, for the first time, by fused deposition modeling (FDM) with isotropic properties has been achieved. Here, bisphenol A-based epoxy and benzoxazine were formulated to a low-temperature thermoplastic and high-temperature thermoset resin, which is melt-extrudable and can be postcured into covalently cross-linked material. Carbon nanotube (CNT) was added in the resin to work as both mechanical enhancement filler and rheology modifier to prevent shape deformation during postcuring process. The cross-layer reaction fuses individual layers into an integrity, thus eliminating layer delamination induced by FDM, offering isotropic mechanical properties regardless of the printing orientations. The highly cross-linked network provides outstanding mechanical strength and superb thermal stability. The excellent shape memory performance with fast recovery rate and large recovery degree is also obtained in the three-dimensional (3D) printed composites.
We discuss the simulation of floating gate MOSFET memory performance consists of Si/Ge/Si quantum dots for electronics storage nodes with higk-κ material as the tunnel oxide. Heterostructure quantum dot was proposed to maintain the good memory performance without losing the long retention characteristic. As the results, retention time could be increase significantly without deteriorating the writing/erasing speed. By selected the dot size of 9–15nm and the tunnel oxide thickness of 2nm, the retention characteristic can reach up to 10{sup 9}s while the writing and erasing time at the order of 10{sup −6}s. By replace the SiO{sub 2} tunnel oxide with high-κ material such as HfO{sub 2}, ZrO{sub 2}, and Y{sub 2}O{sub 3} the leakage current due to the shrinkage of tunnel oxide thickness can be suppressed by the factor of 10 for the EOT lower than ∼1nm. Moreover, the charge-trapping that generated by the defect at Si substrate and higk-κ material interface are fully considered. We found that, the charge-trapping affects the retention time and memory operation performance. By increasing the charge-trapping depth and width, the memory operation time decrease significantly while the retention characteristic slightly improved.
The end of Dennard scaling signaled a shift in HPC supercomputer architectures from systems built from single-core processor architectures to systems built from multicore and eventually manycore architectures. This transition substantially complicated performance optimization and analysis as new programming models were created, new scaling methodologies deployed, and on-chip contention became a bottleneck to performance. Existing distributed memory performance models like logP and logGP were unable to capture this contention. The Roofline model was created to address this contention and its interplay with locality. However, to date, the Roofline model has focused on full-node concurrency. In this paper, we extend the Roofline model to capture the effects of concurrency on data locality and on-chip contention. We demonstrate the value of this new technique by evaluating the NAS parallel benchmarks on both multicore and manycore architectures under both strong-And weak-scaling regimes. In order to quantify the interplay between programming model and locality, we evaluate scaling under both the OpenMP and flat MPI programming models.
The goal of this project is to develop a set of techniques and software tools to enhance the matching between memory accesses in dynamic simulations and the prominent features of modern and future manycore systems, alleviating the memory performance issues for exascale computing. In the first three years, the PI and his group have achieves some significant progress towards the goal, producing a set of novel techniques for improving the memory performance and data locality in manycore systems, yielding 18 conference and workshop papers and 4 journal papers and graduating 6 Ph.Ds. This report summarizes the research results of this project through that period.