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.
The program extends the analysis capabilities of existing profilers through additional analysis to provide insights about communication computation overlap, communication modeling, and broad performance summaries of CPU, GPU and memory. This allows users to understand how their applications are spending time and where performance bottlenecks are.
Modern scientific applications are complicated and require coordination of several components. Proxy application driven software-hardware co-design plays a vital role in driving innovation among the developments of applications, software infrastructure and hardware architecture. Proxy applications are self-contained and simplified codes that are intended to model the performance-critical computations within applications. Applications executing on modern High Performance Computing (HPC) systems are susceptible to network congestion, insufficient memory bandwidth within and across compute nodes, and inadvertent loss of performance due to bugs and unoptimized programming models. Modern numerical simulations and machine learning models play a critical role in studying physical phenomenon under myriad uncertainties. Such applications often exhibit irregular computation and memory accesses at specific regions of the application code, which can contribute to various performance bottlenecks at scale. To mitigate such issues and prepare the next generation hardware for a variety of computation and data movement contingencies, a well-known practice is to consider "proxy" applications as representative motifs for various classes of scientific applications. While there is disagreement in the HPC community on the mechanisms of construction of the proxy applications, there is a strong consensus on their positive impact in co-design. Proxy Applications for Converged Workloads (PACER) is about facilitating software-hardware co-design through proxy applications with the goal of improving the performance of converged science workflows on heterogeneous systems.
The singles and doubles coupled-cluster method that includes a perturbational estimate of connected triple excitations, denoted CCSD(T), has been used, in conjunction with approximate integral techniques, to compute highly accurate rovibrational spectroscopic constants of cyclopropenylidene, C3H2. The approximate integral technique was proposed in 1994 by Rendell and Lee in order to avoid disk storage and input/output bottlenecks, and today it will also significantly aid in the development of algorithms for distributed memory, massively parallel computer architectures. It is shown in this study that use of approximate integrals does not impact the accuracy of CCSD(T) calculations. In addition, the most accurate spectroscopic data yet for C3H2 is presented based on a CCSD(T)/cc-pVQZ quartic force field that is modified to include the effects of core-valence electron correlation. Cyclopropenylidene is of great astronomical and astrobiological interest because it is the smallest aromatic ringed compound to be positively identified in the interstellar medium, and is thus involved in the prebiotic processing of carbon and hydrogen. The singles and doubles coupled-cluster method that includes a perturbational estimate of
As the limits of Moore’s Law approaches, new computing paradigms are developing to breakthrough this bottleneck. One such computer architecture is neuromorphic computing, which models the brain. Electrochemical random-access memory (ECRAM) is a low power and energy efficient memory due to characteristics, such as in-memory compute, ion modulation of the channel conductance, and computation distribution with large scale array integration.
Footprint and reuse distance measure temporal locality and therefore do not capture the significance of access patterns (spacial locality). A strided access pattern has the largest possible footprint but usually has the best performance. To highlight exposed memory latency, we separate footprint into strided (prefetchable) and irregular (non-prefetchable) access components and calculate the growth rate of each. To rapidly compute these footprint access diagnostics, we present two methods, whole-program and precise. Current footprint analyses can cause 200× or more slowdown with realistic inputs and are therefore impractical. Our whole-program method reduces the overhead to 10% by computing upper bounds, but still yields inter-procedural insight through a call path profile. Our precise method uses additional static analysis and profiling to refine the upper bounds for intra-procedural loop nests. We evaluate our approaches using benchmarks that vary access patterns (strided \vs unpredictable), sparsity (all words in a cache line \vs some), and reuse (varying and repeated accesses per element). Notably, for loop nests with unpredictable accesses, the precise method's accuracy is within 10% of ideal. The whole-program method has sufficient accuracy to diagnose bottlenecks.
Conventional tomographic reconstruction typically depends on centralized servers for both data storage and computation, leading to concerns about memory limitations and data privacy. Distributed reconstruction algorithms mitigate these issues by partitioning data across multiple nodes, reducing server load and enhancing privacy. However, these algorithms often encounter challenges related to memory constraints and communication overhead between nodes. In this paper, we introduce a decentralized Alternating Directions Method of Multipliers (ADMM) with configurable quantization. By distributing local objectives across nodes, our approach is highly scalable and can efficiently reconstruct images while adapting to available resources. To overcome communication bottlenecks, we propose two quantization techniques based on K-means clustering and JPEG compression. Numerical experiments with benchmark images illustrate the tradeoffs between communication efficiency, memory use, and reconstruction accuracy.
Vision Transformers (ViTs) are pivotal for foundational models in scientific imagery, including Earth science applications, due to their capability to process large sequence lengths. While transformers for text have inspired scaling sequence lengths in ViTs, adapting these for ViTs introduces unique challenges. We develop distributed sequence parallelism for ViTs, enabling them to handle up to 1M tokens. Our approach, leveraging DeepSpeed-Ulysses and Long-Sequence-Segmentation with model sharding, is the first to apply sequence parallelism in ViT training, achieving a 94% batch scaling efficiency on 2,048 AMD-MI250X GPUs. Evaluating sequence parallelism in ViTs, particularly in models up to 10B parameters, highlighted substantial bottlenecks. We countered these with hybrid sequence, pipeline, and flash attention strategies, to scale beyond single GPU memory limits. Our method significantly enhances climate modeling accuracy by 20% in temperature predictions, marking the first training of a vision transformer model to convergence with a sequence length of 188K tokens, using full self-attention.
Due to the high quantity of operations that spaceborne processing systems must carry out in space, new methodologies and techniques are being presented as good alternatives in order to free the main processor from work and improve the overall performance. These include the development of ancillary dedicated hardware circuits that carry out the more redundant and computationally expensive operations in a faster way, leaving the main processor free to carry out other tasks while waiting for the result. One of these devices is SpaceCube, a FPGA-based system designed by NASA. The opportunity to use FPGA reconfigurable architectures in space allows not only the optimization of the mission operations with hardware-level solutions, but also the ability to create new and improved versions of the circuits, including error corrections, once the satellite is already in orbit. In this work, we propose the optimization of a common operation in remote sensing: the Multi-Spectral Euclidean Distance calculation. For that, two different hardware architectures have been designed and implemented in a Xilinx Virtex-5 FPGA, the same model of FPGAs used by SpaceCube. Previous results have shown that the communications between the embedded processor and the circuit create a bottleneck that affects the overall performance in a negative way. In order to avoid this, advanced methods including memory sharing, Native Port Interface (NPI) connections and Data Burst Transfers have been used.
This exploratory study initiated our inquiry into algorithms and applications that would benefit by latency tolerant approach to algorithm building, including the construction of new algorithms where appropriate. In a multithreaded execution, when a processor reaches a point where remote memory access is necessary, the request is sent out on the network and a context--switch occurs to a new thread of computation. This effectively masks a long and unpredictable latency due to remote loads, thereby providing tolerance to remote access latency. We began to develop standards to profile various algorithm and application parameters, such as the degree of parallelism, granularity, precision, instruction set mix, interprocessor communication, latency etc. These tools will continue to develop and evolve as the Information Power Grid environment matures. To provide a richer context for this research, the project also focused on issues of fault-tolerance and computation migration of numerical algorithms and software. During the initial phase we tried to increase our understanding of the bottlenecks in single processor performance. Our work began by developing an approach for the automatic generation and optimization of numerical software for processors with deep memory hierarchies and pipelined functional units. Based on the results we achieved in this study we are planning to study other architectures of interest, including development of cost models, and developing code generators appropriate to these architectures.
Identifying similar protein sequences is a core step in many computational biology pipelines such as detection of homologous protein sequences, generation of similarity protein graphs for downstream analysis, functional annotation, and gene location. Performance and scalability of protein similarity search have proven to be a bottleneck in many bioinformatics pipelines due to increase in cheap and abundant sequencing data. This work presents a new distributed-memory software PASTIS. PASTIS relies on sparse matrix computations for efficient identification of possibly similar proteins. We use distributed sparse matrices for scalability and show that the sparse matrix infrastructure is a great fit for protein similarity search when coupled with a fully-distributed dictionary of sequences that allow remote sequence requests to be fulfilled. Our algorithm incorporates the unique bias in amino acid sequence substitution in search without altering basic sparse matrix model, and in turn, achieves ideal scaling up to millions of protein sequences.
The use of the general purpose interface bus IEEE-488 to met the special requirements of wind tunnel testing involving PCs is discussed. The gearing of instrumentation to minimize test time, the choice of software to meet computer memory constraints, the use of graphics for improved use of tunnel run time, and the choice of data acquisition equipment to remove bottlenecks are addressed.
We are motivated by newly proposed methods for data mining large-scale corpora of scholarly publications, such as the full biomedical literature, which may consist of tens of millions of papers spanning decades of research. In this setting, analysts seek to discover how concepts relate to one another. They construct graph representations from annotated text databases and then formulate the relationship-mining problem as one of computing all-pairs shortest paths (APSP), which becomes a significant bottleneck. In this context, we present a new high-performance algorithm and implementation of the Floyd-Warshall algorithm for distributed-memory parallel computers accelerated by GPUs, which we call DSNAPSHOT (Distributed Accelerated Semiring All-Pairs Shortest Path). For our largest experiments, we ran DSNAPSHOT on a connected input graph with millions of vertices using 4, 096nodes (24,576GPUs) of the Oak Ridge National Laboratory's Summit supercomputer system. We find DSNAPSHOT achieves a sustained performance of 136×1015 floating-point operations per second (136petaflop/s) at a parallel efficiency of 90% under weak scaling and, in absolute speed, 70% of the best possible performance given our computation (in the single-precision tropical semiring or “min-plus” algebra). Looking forward, we believe this novel capability will enable the mining of scholarly knowledge corpora when embedded and integrated into artificial intelligence-driven natural language processing workflows at scale.
Many critical phenomena in soft matter occur at large length scales, necessitating the resolution of their structure and dynamics at low wavenumbers. However, resolving wavenumber-dependent dynamics computationally via molecular dynamics simulations presents significant challenges, as these phenomena span several orders of magnitude in both time and length scales, resulting in high computational costs and memory demands. Here, this work highlights the computational and memory challenges associated with analyzing molecular trajectories in reciprocal space and demonstrates a method to address them. We introduce FLAMESFast, Low-storage, Accurate, and Memory-Efficient adaptive Sampling, which is a direct method for calculation of structure factors, allowing us to select only the required number of wavevectors for binning. We also use wavenumber-dependent time steps to extract dynamics. Our FLAMES approach effectively mitigates computational and memory/storage bottlenecks. We demonstrate the method using simulations of a model system, liquid octane, at various temperatures. Comparisons with experimental data and real space computation show that the FLAMES technique achieves high accuracy in resolving temperature- and spatially dependent dynamics while being significantly more computationally efficient and requiring less memory and storage than methods based on a uniform wavevector grid and fixed temporal spacing.
We present algorithmic improvements to the loading operations of certain reduced data ensembles produced from neutron scattering experiments at Oak Ridge National Laboratory (ORNL) facilities. Ensembles from multiple measurements are required to cover a wide range of the phase space of a sample material of interest. They are stored using the standard NeXus schema on individual HDF5 files. This makes it a scalability challenge, as the number of experiments stored increases in a single ensemble file. The present work follows up on our previous efforts on data management algorithms, to address identified input output (I/O) bottlenecks in Mantid, an open-source data analysis framework used across several neutron science facilities around the world. We reuse an in-memory binary-tree metadata index that resembles data access patterns, to provide a scalable search and extraction mechanism. In addition, several memory operations are refactored and optimized for the current common use cases, ranging most frequently from 10 to 180, and up to 360 separate measurement configurations. Results from this work show consistent speed ups in wall-clock time on the Mantid LoadMD routine, ranging from 19% to 23% on average, on ORNL production computing systems. The latter depends on the complexity of the targeted instrument-specific data and the system I/O and compute variability for the shared computational resources available to users of ORNL’s Spallation Neutron Source (SNS) and the High Flux Isotope Reactor (HFIR) instruments. Nevertheless, we continue to highlight the need for more research to address reduction challenges as experimental data volumes, user time and processing costs increase.
We report a comparison of two linear-scaling methods which avoid the diagonalization bottleneck of traditional electronic structure algorithms. The Chebyshev expansion method (CEM) is implemented for carbon tight-binding calculations of large systems and its memory and timing requirements compared to those of our previously implemented conjugate gradient density matrix search (CG-DMS). Benchmark calculations are carried out on icosahedral fullerenes from C60 to C8640 and the linear scaling memory and CPU requirements of the CEM demonstrated. We show that the CPU requisites of the CEM and CG-DMS are similar for calculations with comparable accuracy.
MTTKRP is the bottleneck operation in algorithms used to compute the CP tensor decomposition. For sparse tensors, utilizing the compressed sparse fibers (CSF) storage format and the CSF-oriented MTTKRP algorithms is important for both memory and computational efficiency on distributed-memory architectures. Existing intelligent tensor partitioning models assume the computational cost of MTTKRP to be proportional to the total number of nonzeros in the tensor. However, this is not the case for the CSF-oriented MTTKRP on distributed-memory architectures. We outline two deficiencies of nonzero-based intelligent partitioning models when CSF-oriented MTTKRP operations are performed locally: failure to encode processors' computational loads and increase in total computation due to fiber fragmentation. We focus on existing fine-grain hypergraph model and propose a novel vertex weighting scheme that enables this model encode correct computational loads of processors. We also propose to augment the fine-grain model by fiber nets for reducing the increase in total computational load via minimizing fiber fragmentation. In this way, the proposed model encodes minimizing the load of the bottleneck processor. In conclusion, parallel experiments with real-world sparse tensors on up to 1024 processors prove the validity of the outlined deficiencies and demonstrate the merit of our proposed improvements in terms of parallel runtimes.
Abstract The superior size and power scaling potential of ferroelectric-gated Mott transistors makes them promising building blocks for developing energy-efficient memory and logic applications in the post-Moore’s Law era. The close to metallic carrier density in the Mott channel, however, imposes the bottleneck for achieving substantial field effect modulation via a solid-state gate. Previous studies have focused on optimizing the thickness, charge mobility, and carrier density of single-layer correlated channels, which have only led to moderate resistance switching at room temperature. Here, we report a record high nonvolatile resistance switching ratio of 38,440% at 300 K in a prototype Mott transistor consisting of a ferroelectric PbZr 0.2 Ti 0.8 O 3 gate and an R NiO 3 ( R : rare earth)/La 0.67 Sr 0.33 MnO 3 composite channel. The ultrathin La 0.67 Sr 0.33 MnO 3 buffer layer not only tailors the carrier density profile in R NiO 3 through interfacial charge transfer, as corroborated by first-principles calculations, but also provides an extended screening layer that reduces the depolarization effect in the ferroelectric gate. Our study points to an effective material strategy for the functional design of complex oxide heterointerfaces that harnesses the competing roles of charge in field effect screening and ferroelectric depolarization effects.
HipMCL is a high-performance distributed memory implementation of the popular Markov Cluster Algorithm (MCL) and can cluster large-scale networks within hours using a few thousand CPU-equipped nodes. It relies on sparse matrix computations and heavily makes use of the sparse matrix-sparse matrix multiplication kernel (SpGEMM). The existing parallel algorithms in HipMCL are not scalable to Exascale architectures, both due to their communication costs dominating the runtime at large concurrencies and also due to their inability to take advantage of accelerators that are increasingly popular. In this work, we systematically remove scalability and performance bottlenecks of HipMCL. We enable GPUs by performing the expensive expansion phase of the MCL algorithm on GPU. Additionally, we propose a CPU-GPU joint distributed SpGEMM algorithm called pipelined Sparse SUMMA and integrate a probabilistic memory requirement estimator that is fast and accurate. Furthermore, we develop a new merging algorithm for the incremental processing of partial results produced by the GPUs, which improves the overlap efficiency and the peak memory usage. We also integrate a recent and faster algorithm for performing SpGEMM on CPUs. We validate our new algorithms and optimizations with extensive evaluations. With the enabling of the GPUs and integration of new algorithms, HipMCL is up to 12.4x faster, being able to cluster a network with 70 million proteins and 68 billion connections just under 15 minutes using 1024 nodes of ORNL's Summit supercomputer.