Flexible All-Digital Receiver for Bandwidth Efficient Modulations
An all-digital high data rate parallel receiver architecture developed jointly by Goddard Space Flight Center and the Jet Propulsion Laboratory is presented.
Bandwidth↗
SEARCH · Engineering Papers
Results for “Data parallelism”
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.
Lunar Simulant Deposition Technique for Dust Tolerance Studies
A renewed interest in lunar exploration has spawned an array of development efforts for lunar surface assets. These systems depend on the reliable operation of mechanisms and components that may be susceptible to performance degradations or failure due to dust. The Uniform Dust Deposition System was developed at the NASA Glenn Research Center to provide repeatable, uniform, and automated deposition of simulants on surfaces of interest for dust mitigation testing. The system is capable of depositing simulants on test articles up to 60 cm in diameter and 15 cm high in a dry air environment with less than 1 percent relative humidity while keeping users safe from aerosolized dust. The automation of the system allows for high testing throughput while not sacrificing test quality and allows the user to reduce data in parallel. The additional development of a simulant preparation technique complements the repeatability of the deposition physics during testing. The system includes an imaging subsystem that leverages the power of machine learning to count simulant particles and measure their size, thereby allowing for accurate predictions of surface deposition densities (coefficient of determination R^(2) = 0.93) from images alone. The coverage of dust on a surface was shown to be uniform (coefficient of variation CV < 0.11), allowing developers to accurately evaluate the performance of their technology with a prescribed amount of lunar simulant, information that can be used to develop and refine models. The accuracy of the system is currently less than desired for a single deposition run, with a standard deviation (SD) ranging from 18 to 24 mg, or 0.839 to 1.184 mg/sq. cm , for a 5-cm-diameter area. However, the accuracy can be improved by performing multiple deposition runs to build dust to a desired level. Testing has shown that a SD of 0.2 to 0.6 mg, or 0.076 to 0.227 mg/sq. cm, can be achieved for a 5-cm-diameter area using this technique.
Stephen Gerdts↗
Systems and methods for tensor scheduling
A technique for efficient scheduling of operations in a program for parallelized execution thereof using a multi-processor runtime environment having two or more processors includes constraining the type or number of loop optimization transforms that may be explored such that memory and processing capacity available for the scheduling task are not exceeded, while facilitating a tradeoff between memory locality, parallelization, and/or data communication between memory modules of the multi-processor runtime environment.
Meister, Benoit J.↗
Generalizing mkFit and its Application to HL-LHC
mkFit is an implementation of the Kalman filter-based track reconstruction algorithm that exploits both thread- and data-level parallelism. In the past few years the project transitioned from the R&D phase to deployment in the Run-3 offline workflow of the CMS experiment. The CMS tracking performs a series of iterations, targeting reconstruction of tracks of increasing difficulty after removing hits associated to tracks found in previous iterations. mkFit has been adopted for several of the tracking iterations, which contribute to the majority of reconstructed tracks. When tested in the standard conditions for production jobs, speedups in track pattern recognition are on average of the order of 3.5x for the iterations where it is used (3-7x depending on the iteration). Multiple factors contribute to the observed speedups, including vectorization and a lightweight geometry description, as well as improved memory management and single precision. Efficient vectorization is achieved with both the icc and the gcc (default in CMSSW) compilers and relies on a dedicated library for small matrix operations, Matriplex, which has recently been released in a public repository. While the mkFit geometry description already featured levels of abstraction from the actual Phase-1 CMS tracker, several components of the implementations were still tied to that specific geometry. We have further generalized the geometry description and the configuration of the run-time parameters, in order to enable support for the Phase-2 upgraded tracker geometry for the HL-LHC and potentially other detector configurations. The implementation strategy and high-level code changes required for the HL-LHC geometry are presented. Speedups in track building from mkFit imply that track fitting becomes a comparably time consuming step of the tracking chain.
72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
HARMONY: Large-Scale Architecture Search for Efficient Hybrid Language Models
As large language models scale to trillions of parameters, their computational and memory requirements present critical challenges for efficient training and deployment. While Mixture of Experts (MoE) architectures enable efficient scaling through sparse parameter activation, and state-space models like Mamba offer linear-time complexity, principled methods for combining these paradigms remain undeveloped. We introduce HARMONY (Hybrid Architecture Research for Mamba, Optimized with Neural efficiencY), a multi-objective evolutionary neural architecture search framework for discovering efficient hybrid language models that integrate Transformer attention mechanisms, Mixture-of-Experts routing, and Mamba state-space components. Through large-scale distributed search using 16,384 MI250X GPUs on the Frontier supercomputer, HARMONY explores a comprehensive design space encompassing six attention variants (MHA, MQA, GQA, MLA, SWA, and Mamba-2), variable MoE configurations with both routed and shared experts, and extensive Mamba hyperparameters. Our framework discovers heterogeneous architectures that balance training performance with computational efficiency through multi-objective optimization incorporating latency penalties and fitness-based selection. Analysis of discovered architectures reveals that optimal hybrid designs favor heterogeneous component mixing rather than homogeneous patterns, with Mamba-2 and Multi-Head Latent Attention (MLA) emerging as preferred mechanisms. Discovered architectures demonstrate superior training efficiency: our best configuration achieves a final perplexity of 1.0874 with 2.38B parameters while processing 4,320 tokens/second, outperforming significantly larger manually designed models. Full-scale evaluation shows HARMONY's top architectures achieve better loss trajectories than equivalently-sized models using state-of-the-art configurations including Mixtral, Jamba, and Samba. Additionally, we demonstrate 91% weak scaling efficiency when training discovered 36B-parameter models across 1,024 GPUs. HARMONY is released as an open framework with comprehensive tools for building and training hybrid models using expert-data-pipeline parallelism, democratizing access to automated architecture design for next-generation language models.
Herron, Emily [ORNL] (ORCID:0000000273008172)↗
Runtime volume visualization for parallel CFD
This paper discusses some aspects of design of a data distributed, massively parallel volume rendering library for runtime visualization of parallel computational fluid dynamics simulations in a message-passing environment. Unlike the traditional scheme in which visualization is a postprocessing step, the rendering is done in place on each node processor. Computational scientists who run large-scale simulations on a massively parallel computer can thus perform interactive monitoring of their simulations. The current library provides an interface to handle volume data on rectilinear grids. The same design principles can be generalized to handle other types of grids. For demonstration, we run a parallel Navier-Stokes solver making use of this rendering library on the Intel Paragon XP/S. The interactive visual response achieved is found to be very useful. Performance studies show that the parallel rendering process is scalable with the size of the simulation as well as with the parallel computer.
Ma, Kwan-Liu↗
Implementation of Helioseismic Data Reduction and Diagnostic Techniques on Massively Parallel Architectures
Under the direction of Dr. Rhodes, and the technical supervision of Dr. Korzennik, the data assimilation of high spatial resolution solar dopplergrams has been carried out throughout the program on the Intel Delta Touchstone supercomputer. With the help of a research assistant, partially supported by this grant, and under the supervision of Dr. Korzennik, code development was carried out at SAO, using various available resources. To ensure cross-platform portability, PVM was selected as the message passing library. A parallel implementation of power spectra computation for helioseismology data reduction, using PVM was successfully completed. It was successfully ported to SMP architectures (i.e. SUN), and to some MPP architectures (i.e. the CM5). Due to limitation of the implementation of PVM on the Cray T3D, the port to that architecture was not completed at the time.
Korzennik, Sylvain↗
Refining image segmentation by integration of edge and region data
An iterative parallel region growing (IPRG) algorithm previously developed by Tilton (1989) produces hierarchical segmentations of images from finer to coarser resolution. An ideal segmentation does not always correspond to one single iteration but to several different ones, each one producing the 'best' result for a separate part of the image. With the goal of finding this ideal segmentation, the results of the IPRG algorithm are refined by utilizing some additional information, such as edge features, and by interpreting the tree of hierarchical regions.
Le Moigne, Jacqueline↗
Technology Assessment of High Capacity Data Storage Systems: Can We Avoid a Data Survivability Crisis?
This technology assessment of long-term high capacity data storage systems identifies an emerging crisis of severe proportions related to preserving important historical data in science, healthcare, manufacturing, finance and other fields. For the last 50 years, the information revolution, which has engulfed all major institutions of modem society, centered itself on data-their collection, storage, retrieval, transmission, analysis and presentation. The transformation of long term historical data records into information concepts, according to Drucker, is the next stage in this revolution towards building the new information based scientific and business foundations. For this to occur, data survivability, reliability and evolvability of long term storage media and systems pose formidable technological challenges. Unlike the Y2K problem, where the clock is ticking and a crisis is set to go off at a specific time, large capacity data storage repositories face a crisis similar to the social security system in that the seriousness of the problem emerges after a decade or two. The essence of the storage crisis is as follows: since it could take a decade to migrate a peta-byte of data to a new media for preservation, and the life expectancy of the storage media itself is only a decade, then it may not be possible to complete the transfer before an irrecoverable data loss occurs. Over the last two decades, a number of anecdotal crises have occurred where vital scientific and business data were lost or would have been lost if not for major expenditures of resources and funds to save this data, much like what is happening today to solve the Y2K problem. A pr-ime example was the joint NASA/NSF/NOAA effort to rescue eight years worth of TOVS/AVHRR data from an obsolete system, which otherwise would have not resulted in the valuable 20-year long satellite record of global warming. Current storage systems solutions to long-term data survivability rest on scalable architectures having parallel paths for data migration.
Halem, M.↗
Gravitational Waves: Search Results, Data Analysis and Parameter Estimation. Amaldi 10 Parallel Session C2
The Amaldi 10 Parallel Session C2 on gravitational wave(GW) search results, data analysis and parameter estimation included three lively sessions of lectures by 13 presenters, and 34 posters. The talks and posters covered a huge range of material, including results and analysis techniques for ground-based GW detectors, targeting anticipated signals from different astrophysical sources: compact binary inspiral, merger and ringdown; GW bursts from intermediate mass binary black hole mergers, cosmic string cusps, core-collapse supernovae, and other unmodeled sources; continuous waves from spinning neutron stars; and a stochastic GW background. There was considerable emphasis on Bayesian techniques for estimating the parameters of coalescing compact binary systems from the gravitational waveforms extracted from the data from the advanced detector network. This included methods to distinguish deviations of the signals from what is expected in the context of General Relativity.
Tests of general relativity↗
Database Reorganization in Parallel Disk Arrays with I/O Service Stealing
We present a model for data reorganization in parallel disk systems that is geared towards load balancing in an environment with periodic access patterns. Data reorganization is performed by disk cooling, i.e. migrating files or extents from the hottest disks to the coldest ones. We develop an approximate queueing model for determining the effective arrival rates of cooling requests and discuss its use in assessing the costs versus benefits of cooling.
Zabback, Peter↗
Image segmentation by iterative parallel region growing with application to data compression and image analysis
Image segmentation can be a key step in data compression and image analysis. However, the segmentation results produced by most previous approaches to region growing are suspect because they depend on the order in which portions of the image are processed. An iterative parallel segmentation algorithm avoids this problem by performing globally best merges first. Such a segmentation approach, and two implementations of the approach on NASA's Massively Parallel Processor (MPP) are described. Application of the segmentation approach to data compression and image analysis is then described, and results of such application are given for a LANDSAT Thematic Mapper image.
Tilton, James C.↗
A Parallel Pipelined Renderer for the Time-Varying Volume Data
This paper presents a strategy for efficiently rendering time-varying volume data sets on a distributed-memory parallel computer. Time-varying volume data take large storage space and visualizing them requires reading large files continuously or periodically throughout the course of the visualization process. Instead of using all the processors to collectively render one volume at a time, a pipelined rendering process is formed by partitioning processors into groups to render multiple volumes concurrently. In this way, the overall rendering time may be greatly reduced because the pipelined rendering tasks are overlapped with the I/O required to load each volume into a group of processors; moreover, parallelization overhead may be reduced as a result of partitioning the processors. We modify an existing parallel volume renderer to exploit various levels of rendering parallelism and to study how the partitioning of processors may lead to optimal rendering performance. Two factors which are important to the overall execution time are re-source utilization efficiency and pipeline startup latency. The optimal partitioning configuration is the one that balances these two factors. Tests on Intel Paragon computers show that in general optimal partitionings do exist for a given rendering task and result in 40-50% saving in overall rendering time.
Chiueh, Tzi-Cker↗
High Rate Digital Demodulator ASIC
The architecture of High Rate (600 Mega-bits per second) Digital Demodulator (HRDD) ASIC capable of demodulating BPSK and QPSK modulated data is presented in this paper. The advantages of all-digital processing include increased flexibility and reliability with reduced reproduction costs. Conventional serial digital processing would require high processing rates necessitating a hardware implementation in other than CMOS technology such as Gallium Arsenide (GaAs) which has high cost and power requirements. It is more desirable to use CMOS technology with its lower power requirements and higher gate density. However, digital demodulation of high data rates in CMOS requires parallel algorithms to process the sampled data at a rate lower than the data rate. The parallel processing algorithms described here were developed jointly by NASA's Goddard Space Flight Center (GSFC) and the Jet Propulsion Laboratory (JPL). The resulting all-digital receiver has the capability to demodulate BPSK, QPSK, OQPSK, and DQPSK at data rates in excess of 300 Mega-bits per second (Mbps) per channel. This paper will provide an overview of the parallel architecture and features of the HRDR ASIC. In addition, this paper will provide an over-view of the implementation of the hardware architectures used to create flexibility over conventional high rate analog or hybrid receivers. This flexibility includes a wide range of data rates, modulation schemes, and operating environments. In conclusion it will be shown how this high rate digital demodulator can be used with an off-the-shelf A/D and a flexible analog front end, both of which are numerically computer controlled, to produce a very flexible, low cost high rate digital receiver.
Ghuman, Parminder↗
Parallelizing the Unpacking and Clustering of Detector Data for Reconstruction of Charged Particle Tracks on Multi-core CPUs and Many-core GPUs
We present results from parallelizing the unpacking and clustering steps of the raw data from the silicon strip modules for reconstruction of charged particle tracks. Throughput is further improved by concurrently processing multiple events using nested OpenMP parallelism on CPU or CUDA streams on GPU. The new implementation along with earlier work in developing a parallelized and vectorized implementation of the combinatoric Kalman filter algorithm has enabled efficient global reconstruction of the entire event on modern computer architectures. We demonstrate the performance of the new implementation on Intel Xeon and NVIDIA GPU architectures.
72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
Monitoring Data-Structure Evolution in Distributed Message-Passing Programs
Monitoring the evolution of data structures in parallel and distributed programs, is critical for debugging its semantics and performance. However, the current state-of-art in tracking and presenting data-structure information on parallel and distributed environments is cumbersome and does not scale. In this paper we present a methodology that automatically tracks memory bindings (not the actual contents) of static and dynamic data-structures of message-passing C programs, using PVM. With the help of a number of examples we show that in addition to determining the impact of memory allocation overheads on program performance, graphical views can help in debugging the semantics of program execution. Scalable animations of virtual address bindings of source-level data-structures are used for debugging the semantics of parallel programs across all processors. In conjunction with light-weight core-files, this technique can be used to complement traditional debuggers on single processors. Detailed information (such as data-structure contents), on specific nodes, can be determined using traditional debuggers after the data structure evolution leading to the semantic error is observed graphically.
Sarukkai, Sekhar R.↗
The impact of satellite soundings on the National Meteorological Center's Analysis and Forecast System - The data systems test results
Parallel analyses produced with (SAT) and without (NOSAT) the sounding data from the experimental Nimbus-6 and operational NOAA-4 satellites for two data system test periods have indicated that satellite soundings had little impact on forecast through 72 h over the Northern Hemisphere. The effect of remote temperature sounding was generally small and of inconsistent sign, i.e. beneficial in some cases, harmful in others. These positive and negative contributions averaged slightly positive for test period 6 and slightly negative for test period 5. Because of the inconsistency in sign, neither result is much of meteorological consequence. There were, however, systematic differences between the SAT and NOSAT mode analysis: the amplitude of weather systems was consistently less in the SAT mode. The reduced amplitude reflects a tendency for the satellite temperature retrievals to underestimate the spatial variance in the thermal structure of the atmosphere.
Tracton, M. S.↗
On the impact of communication complexity in the design of parallel numerical algorithms
This paper describes two models of the cost of data movement in parallel numerical algorithms. One model is a generalization of an approach due to Hockney, and is suitable for shared memory multiprocessors where each processor has vector capabilities. The other model is applicable to highly parallel nonshared memory MIMD systems. In the second model, algorithm performance is characterized in terms of the communication network design. Techniques used in VLSI complexity theory are also brought in, and algorithm independent upper bounds on system performance are derived for several problems that are important to scientific computation.
Gannon, D.↗