Search NASA⌕ Search

SEARCH · Search NASA

Results for “Parallelization Strategy”

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 991 records · Page 55
All sourcesNASA NTRSDOE OSTI
DOE OSTI2014

DIRAQ: scalable in situ data- and resource-aware indexing for optimized query performance

Scientific data analytics in high-performance computing environments has been evolving along with the advancement of computing capabilities. With the onset of exascale computing, the increasing gap between compute performance and I/O bandwidth has rendered the traditional post-simulation processing a tedious process. Despite the challenges due to increased data production, there exists an opportunity to benefit from "cheap" computing power to perform query-driven exploration and visualization during simulation time. To accelerate such analyses, applications traditionally augment, post-simulation, raw data with large indexes, which are then repeatedly utilized for data exploration. However, the generation of current state-of-the-art indexes involves a compute- and memory-intensive processing, thus rendering them inapplicable in an in situ context. In this paper we propose DIRAQ, a parallel in situ, in network data encoding and reorganization technique that enables the transformation of simulation output into a query-efficient form, with negligible runtime overhead to the simulation run. DIRAQ's effective core-local, precision-based encoding approach incorporates an embedded compressed index that is 3---6 $$\times $$ ? smaller than current state-of-the-art indexing schemes. Its data-aware index adjustmentation improves performance of group-level index layout creation by up to 35 % and reduces the size of the generated index by up to 27 %. Moreover, DIRAQ's in network index merging strategy enables the creation of aggregated indexes that speed up spatial-context query responses by up to $$10\times $$ 10 ? versus alternative techniques. DIRAQ's topology-, data-, and memory-aware aggregation strategy results in efficient I/O and yields overall end-to-end encoding and I/O time that is less than that required to write the raw data with MPI collective I/O.

Computer Science↗
DOE OSTI2018

Structure–Property Relationships of Oligonucleotide Polyelectrolyte Complex Micelles

Polyelectrolyte complex micelles (PCMs), nanoparticles formed by electrostatic self-assembly of charged polymers with charged-neutral hydrophilic block copolymers, offer a potential solution to the challenging problem of delivering therapeutic nucleic acids into cells and organisms. Promising results have been reported in vitro and in animal models but basic structure property relationships are largely lacking, and some reports have suggested that double stranded nucleic acids cannot form PCMs due to their high bending rigidity. Here, we report on a study of PCMs formed by DNA oligonucleotides of varied length and hybridization state and poly(L)lysine-poly(ethylene glycol) block copolymers with varying block lengths. We employ a multimodal characterization strategy combining small-angle X-ray scattering (SAXS), multiangle light scattering (MALS), and cryo-electron microscopy (cryo-TEM) to simultaneously probe the morphology and internal structure of the micelles. Over a wide range of parameters, we find that nanoparticle shape is controlled primarily by the hybridization state of the oligonucleotides with single-stranded oligonucleotides forming spheroidal micelles and double stranded oligonucleotides forming wormlike micelles. The length of the charged block controls the radius of the nanoparticle, while oligonucleotide length appears to have little impact on either size or shape. At smaller length scales, we observe parallel packing of DNA helices inside the double-stranded nanoparticles, consistent with results from condensed genomic DNA. We additionally describe salt- and thermal-annealing protocols for preparing PCMs with high repeatability and low polydispersity. Together, these results provide a capability to rationally design PCMs with desired sizes and shapes that should greatly assist development of this promising delivery technology.

77 NANOSCIENCE AND NANOTECHNOLOGY↗
DOE OSTI2012

Initiation stress threshold irradiation assisted stress corrosion cracking criterion assessment for core internals in PWR environment

Irradiation assisted stress corrosion cracking (IASCC) is a problem of growing importance in pressurized water reactors (PWR). An understanding of the mechanism(s) of IASCC is required in order to provide guidance for the development of mitigation strategies. One of the principal reasons why the IASCC mechanism(s) has been so difficult to understand is the inseparability of the different IASCC potential contributors evolutions due to neutron irradiation. The potential contributors to IASCC in PWR primary water are: (i) radiation induced segregation (RIS) at grain boundaries, (ii) radiation induced microstructure (formation and growth of dislocations loops, voids, bubbles, phases), (iii) localized deformation under loading, (iv) irradiation creep and transmutations. While the development of some of the contributors (RIS, microstructure) with increasing doses are at least qualitatively well understood, the role of these changes on IASCC remains unclear. Parallel to fundamental understanding developments relative to IASCC, well controlled laboratory tests on neutron irradiated stainless steels are needed to assess the main mechanisms and also to establish an engineering criterion relative to the initiation of fracture due to IASCC. First part of this study describes the methodology carried out at CEA in order to provide more experimental data from constant load tests dedicated to the study of initiation of SCC on neutron irradiated stainless steel. A description of the autoclave recirculation loop dedicated to SCC tests on neutron irradiated materials is then given. This autoclave recirculation loop has been started on July 2010 with the first SCC test on an irradiated stainless steel (grade 316) performed at CEA. The main steps of the interrupted SCC tests are then described. Second part of this paper reports the partial results of the first test performed on a highly neutron irradiated material. (authors)

36 MATERIALS SCIENCE↗
DOE OSTI2025

Scale-up Unlearnable Examples Learning with High-performance Computing

Recent advancements in AI models, like ChatGPT, are structured to retain user interactions, which could inadvertently include sensitive healthcare data. In the healthcare field, particularly when radiologists use AI-driven diagnostic tools hosted on online platforms, there is a risk that medical imaging data may be repurposed for future AI training without explicit consent, spotlighting critical privacy and intellectual property concerns around healthcare data usage. Addressing these privacy challenges, a novel approach known as Unlearnable Examples (UEs) has been introduced, aiming to make data unlearnable to deep learning models. A prominent method within this area, called Unlearnable Clustering (UC), has shown improved UE performance with larger batch sizes but was previously limited by computational resources (e.g., a single workstation). To push the boundaries of UE performance with theoretically unlimited resources, we scaled up UC learning across various datasets using Distributed Data Parallel (DDP) training on the Summit supercomputer. Our goal was to examine UE efficacy at high-performance computing (HPC) levels to prevent unauthorized learning and enhance data security, particularly exploring the impact of batch size on UE’s unlearnability. Utilizing the robust computational capabilities of the Summit, extensive experiments were conducted on diverse datasets such as Pets, MedMNist, Flowers, and Flowers102. Our findings reveal that both overly large and overly small batch sizes can lead to performance instability and affect accuracy. However, the relationship between batch size and unlearnability varied across datasets, highlighting the necessity for tailored batch size strategies to achieve optimal data protection. The use of Summit’s high-performance GPUs, along with the efficiency of the DDP framework, facilitated rapid updates of model parameters and consistent training across nodes. Our results underscore the critical role of selecting appropriate batch sizes based on the specific characteristics of each dataset to prevent learning and ensure data security in deep learning applications. The source code is publicly available at https: // github. com/ hrlblab/ UE_ HPC .

Zhu, Yanfan [Vanderbilt University, Nashville, TN,↗
DOE OSTI2015

Parallel implementation and performance optimization of the configuration-interaction method

The configuration-interaction (CI) method, long a popular approach to describe quantum many-body systems, is cast as a very large sparse matrix eigenpair problem with matrices whose dimension can exceed one billion. Such formulations place high demands on memory capacity and memory bandwidth - - two quantities at a premium today. In this paper, we describe an efficient, scalable implementation, BIGSTICK, which, by factorizing both the basis and the interaction into two levels, can reconstruct the nonzero matrix elements on the fly, reduce the memory requirements by one or two orders of magnitude, and enable researchers to trade reduced resources for increased computational time. We optimize BIGSTICK on two leading HPC platforms - - the Cray XC30 and the IBM Blue Gene/Q. Specifically, we not only develop an empirically-driven load balancing strategy that can evenly distribute the matrix-vector multiplication across 256K threads, we also developed techniques that improve the performance of the Lanczos reorthogonalization. Combined, these optimizations improved performance by 1.3-8× depending on platform and configuration.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS2014

NASA Space Launch System Operations Outlook

The National Aeronautics and Space Administration's (NASA) Space Launch System (SLS) Program, managed at the Marshall Space Flight Center (MSFC), is working with the Ground Systems Development and Operations (GSDO) Program, based at the Kennedy Space Center (KSC), to deliver a new safe, affordable, and sustainable capability for human and scientific exploration beyond Earth's orbit (BEO). Larger than the Saturn V Moon rocket, SLS will provide 10 percent more thrust at liftoff in its initial 70 metric ton (t) configuration and 20 percent more in its evolved 130-t configuration. The primary mission of the SLS rocket will be to launch astronauts to deep space destinations in the Orion Multi- Purpose Crew Vehicle (MPCV), also in development and managed by the Johnson Space Center. Several high-priority science missions also may benefit from the increased payload volume and reduced trip times offered by this powerful, versatile rocket. Reducing the lifecycle costs for NASA's space transportation flagship will maximize the exploration and scientific discovery returned from the taxpayer's investment. To that end, decisions made during development of SLS and associated systems will impact the nation's space exploration capabilities for decades. This paper will provide an update to the operations strategy presented at SpaceOps 2012. It will focus on: 1) Preparations to streamline the processing flow and infrastructure needed to produce and launch the world's largest rocket (i.e., through incorporation and modification of proven, heritage systems into the vehicle and ground systems); 2) Implementation of a lean approach to reach-back support of hardware manufacturing, green-run testing, and launch site processing and activities; and 3) Partnering between the vehicle design and operations communities on state-of-the-art predictive operations analysis techniques. An example of innovation is testing the integrated vehicle at the processing facility in parallel, rather than sequentially, saving both time and money. These themes are accomplished under the context of a new cross-program integration model that emphasizes peer-to-peer accountability and collaboration towards a common, shared goal. Utilizing the lessons learned through 50 years of human space flight experience, SLS is assigning the right number of people from appropriate backgrounds, providing them the right tools, and exercising the right processes for the job. The result will be a powerful, versatile, and capable heavy-lift, human-rated asset for the future human and scientific exploration of space.

Hefner, William Keith↗
NASA NTRS2015

Affordable Development and Demonstration of a Small NTR Engine and Stage: How Small is Big Enough?

The Nuclear Thermal Rocket (NTR) derives its energy from fission of uranium-235 atoms contained within fuel elements that comprise the engine's reactor core. It generates high thrust and has a specific impulse potential of approximately 900 seconds - a 100% increase over today's best chemical rockets. The Nuclear Thermal Propulsion (NTP) project, funded by NASA's AES program, includes five key task activities: (1) Recapture, demonstration, and validation of heritage graphite composite (GC) fuel (selected as the "Lead Fuel" option); (2) Engine Conceptual Design; (3) Operating Requirements Definition; (4) Identification of Affordable Options for Ground Testing; and (5) Formulation of an Affordable Development Strategy. During FY'14, a preliminary DDT&E plan and schedule for NTP development was outlined by GRC, DOE and industry that involved significant system-level demonstration projects that included GTD tests at the NNSS, followed by a FTD mission. To reduce cost for the GTD tests and FTD mission, small NTR engines, in either the 7.5 or 16.5 klbf thrust class, were considered. Both engine options used GC fuel and a "common" fuel element (FE) design. The small approximately 7.5 klbf "criticality-limited" engine produces approximately 157 megawatts of thermal power (MWt) and its core is configured with parallel rows of hexagonal-shaped FEs and tie tubes (TTs) with a FE to TT ratio of approximately 1:1. The larger approximately 16.5 klbf Small Nuclear Rocket Engine (SNRE), developed by LANL at the end of the Rover program, produces approximately 367 MWt and has a FE to TT ratio of approximately 2:1. Although both engines use a common 35 inch (approximately 89 cm) long FE, the SNRE's larger diameter core contains approximately 300 more FEs needed to produce an additional 210 MWt of power. To reduce the cost of the FTD mission, a simple "1-burn" lunar flyby mission was considered to reduce the LH2 propellant loading, the stage size and complexity. Use of existing and flight proven liquid rocket and stage hardware (e.g., from the RL10B-2 engine and Delta Cryogenic Second Stage) was also maximized to further aid affordability. This paper examines the pros and cons of using these two small engine options, including their potential to support future human exploration missions to the Moon, near Earth asteroids, and Mars, and recommends a preferred size. It also provides a preliminary assessment of the key activities, development options, and schedule required to affordably build, ground test and fly a small NTR engine and stage within a 10-year timeframe.

Spacecraft design↗
NASA NTRSDate not supplied

Recent Advancements in the PATO Material Response Code

Introduction: Predicting the complicated multiphysics phenomena during atmospheric entry requires high-fidelity modeling tools to refine estimates of mission risks during entry. To this end, new capabilities are being added to the Porous-material Analysis Toolbox based on OpenFOAM (PATO) [1,2,3]. PATO is an open-source software for Computational Material Response (CMR) of reactive porous materials submitted to high-temperature environments. The objective of this work is to highlight current efforts to add to and improve upon the modeling capabilities of PATO. These include efforts to loosely couple PATO with other discipline specialized codes including hypersonic Computational Fluid Dynamics (CFD), to assess the interaction effects between pyrolysis gas blowing and the boundary layer, and Computational Solid Mechanics (CSM), to address modeling of mechanical erosion. Other refinements include surface phenomena modeling capabilities to address the effects of silicone-based coatings applied to the TPS during flight preparation, and a unified multiphase solver for a mixed porous-material and plain-fluid domain. Coupling CMR with CFD (CMR/CFD): A loose coupling between PATO and the Data Parallel Line Relaxation (DPLR) [4] CFD code has been achieved by making use of a blowing boundary condition at the heatshield surface available in DPLR. Starting with heat flux estimates with no pyrolysis gas blowing at the surface, blowing gases are computed by the CMR and passed to the CFD such that aerothermal properties of the environment can be recomputed for a new CMR computation. This leads to an iterative process which is supplemented with an estimate of the radiative heat flux using the Nonequilibrium air radiation (NEQAIR) [5] program. The entire iterative process is illustrated in Figure 1. This coupling strategy has been utilized in computing the MSL material response. The goal is to compare the coupled CMR/CFD results with material response results obtained using traditional blowing corrections. Coupling CMS with CMR: A mechanical erosion model is currently being implemented in PATO to account for the additional mass removal induced by high shear conditions. The modeling process at each timestep consists of updating the mechanical properties as a function of temperature and computing the stress tensor and displacement fields of the material. Then, a failure criteria model determines the regions in which the stress exceeds the ultimate strength values resulting in mesh movement to account for mass removal. This model allows the material response simulation to compute the recession due to both oxidation and shear-induced erosion. The model is demonstrated by computing material response of sphere-cone arc jet samples. Surface Modeling Capabilities: NuSil, a silicone-based coating, was sprayed onto the MSL and Mars 2020 heatshields to mitigate shedding of phenolic dust. To better understand the effects of the NuSil coating on the material response, a novel model has been implemented in PATO. In this model, the equilibrium of the charred NuSil surface is modeled as pure silica, and a constant offset, inspired by the classical spallation model, is added to the the char blowing rate and wall enthalpy to reproduce HyMETS experimental results. The model has also been used to estimate the 3D material response of the MSL heatshield [6]. Unified Solver: In addition to the iterative loose coupling approach mentioned above, a multiphase unified solver is being developed to couple the environment (plain-fluid phase) and the porous-material phase. The solver is based on the volume averaged conservation of mass, momentum, and energy for the macroscale with closure models which include microscale effects through effective physicochemical properties. The unified solver has been used to compute flow through a porous plug and solve the Beavers and Joseph problem [7]. Since the strong coupling between phases is inherent to this solver, modeling assumptions present in other coupling methods of material response are mitigated. This strategy also makes it feasible to capture the competition between surface and volume ablation in the same computational domain, which is usually not possible with other coupling approaches.

Thermal Protection Systems↗
NASA NTRS1993

Real-time dynamics and control strategies for space operations of flexible structures

This project (NAG9-574) was meant to be a three-year research project. However, due to NASA's reorganizations during 1992, the project was funded only for one year. Accordingly, every effort was made to make the present final report as if the project was meant to be for one-year duration. Originally, during the first year we were planning to accomplish the following: we were to start with a three dimensional flexible manipulator beam with articulated joints and with a linear control-based controller applied at the joints; using this simple example, we were to design the software systems requirements for real-time processing, introduce the streamlining of various computational algorithms, perform the necessary reorganization of the partitioned simulation procedures, and assess the potential speed-up realization of the solution process by parallel computations. The three reports included as part of the final report address: the streamlining of various computational algorithms; the necessary reorganization of the partitioned simulation procedures, in particular the observer models; and an initial attempt of reconfiguring the flexible space structures.

Park, K. C.↗
DOE OSTI2025

Leveraging the Run 3 experience for the evolution of the ATLAS software-based readout towards HL-LHC

The High-Luminosity Large Hadron Collider (HL-LHC), scheduled to start operating in 2030, aims to increase the instantaneous luminosity by a factor of 10 compared to the LHC. To match this increase, the ATLAS experiment has been implementing a major upgrade program divided into two phases. The first phase (Phase-I), completed in 2022, introduced new trigger and detector systems that have been used during the Run 3 data taking period which began in July 2022. These systems have been used in conjunction with the new Data Acquisition (DAQ) Readout system, based on a software application called Software Readout Driver (SW ROD). SW ROD receives and aggregates data from the front-end electronics via the Front-End Link eXchange (FELIX) system and passes aggregated data fragments to the High-Level Trigger (HLT) system. During Run 3, SW ROD operates in parallel with the legacy Readout System (ROS) at an input rate of 100 kHz. For the Phase-II, the legacy ROS will be completely replaced with a new system based on the next generation of FELIX and an evolution of the SW ROD application called Data Handler. Data Handler has the same functional requirements as SW ROD but must be able to operate at an input rate of 1 MHz. To facilitate this evolution the SW ROD has been implemented using plugin architecture. This contribution presents the design and implementation of the SW ROD application for Run 3, along with the strategy for its evolution to the Phase-II Readout system. It discusses the lessons learned during Run 3 and describes the challenges that have been addressed to accomplish the demanding performance requirements of HL-LHC.

Kolos, Serguei [Univ. of California, Irvine, CA (U↗
NASA NTRS1994

Aligning parallel arrays to reduce communication

Axis and stride alignment is an important optimization in compiling data-parallel programs for distributed-memory machines. We previously developed an optimal algorithm for aligning array expressions. Here, we examine alignment for more general program graphs. We show that optimal alignment is NP-complete in this setting, so we study heuristic methods. This paper makes two contributions. First, we show how local graph transformations can reduce the size of the problem significantly without changing the best solution. This allows more complex and effective heuristics to be used. Second, we give a heuristic that can explore the space of possible solutions in a number of ways. We show that some of these strategies can give better solutions than a simple greedy approach proposed earlier. Our algorithms have been implemented; we present experimental results showing their effect on the performance of some example programs running on the CM-5.

Sheffler, Thomas J.↗
NASA NTRS1993

Relationships between pigment composition variation and reflectance for plant species from a coastal savannah in California

Advances in imaging spectroscopy have indicated that remotely sensed reflectance measurements of the plant canopy may be used to identify and qualify some classes of canopy biochemicals; however, the manner in which differences in biochemical compositions translate into differences is not well understood. Most frequently, multiple linear regression routines have been used to correlate narrow band reflectance values with measured biochemical concentrations. Although some success has been achieved with such methods for given data sets, the bands selected by multiple regression are not consistent between data sets, nor is it always clear what physical or biological basis underlies the correlation. To examine the relationship between biochemical concentration and leaf reflectance signal we chose to focus on the visible spectrum where the primary biochemical absorbances are due to photosynthetic pigments. Pigments provide a range of absorbance features, occur over a range of concentrations in natural samples, and are ecophysiologically important. Concentrations of chlorophyll, for example, have been strongly correlated to foliar nitrogen levels within a species and to photosynthetic capacity across many species. In addition pigments effectively absorb most of the photosynthetically active radiation between 400-700 nm, a spectral region for which silicon detectors have good signal/noise characteristics. Our strategy has been to sample a variety of naturally occurring species to measure leaf reflectance and pigment compositions. We hope to extend our understanding of pigment reflectance effects to interpret small overlapping absorbances of other biochemicals in the infrared region. For this reason, selected samples were also tested to determine total nitrogen, crude protein, cellulose, and lignin levels. Leaf reflectance spectra measured with AVIRIS bandwidths and wavelengths were compared between species and within species and for differences between seasons, for changes in the the shape of the spectra. We attempt to statistically correlate these shape changes with differences in pigment compositions. In parallel with our comparisons of pigment composition and leaf reflectance, we have modified the PROSPECT leaf reflectance model to test the contributions of pigments or pigment group concentrations. PROSPECT considers a leaf as a multi-layer dielectric plane with an uneven surface. Jacquemoud adapted the basic analysis of Allen for surface effects, a leaf thickness factor, and the absorption of water and chlorophyll (actually all pigments) and the plant matrix. Our modifications to PROSPECT in the forward direction include breaking out the pigment concentration parameter into separate components for chlorophyll a and b and a number of xanthophylls and carotenes, and introducing a shift and convolution function to model the spread and shift from their in vitro measurements to their in vivo state. Further, we have considered how the matrix elements (i.e., all biochemicals and structural effects not modeled explicity) vary with species.

Ustin, Susan L.↗
DOE OSTI2022

Data Mining and Visualization of High-Dimensional ICME Data for Additive Manufacturing

Integrated computational materials engineering (ICME) methods combining CALPHAD with process-based simulations can produce rich, high-dimensional data for alloy and process design. In ICME methods for metallurgical applications, the visualization and interpretation of such high-dimensional data has previously been through heat maps represented in 2 or 3 dimensions. While such an approach is ideal when one variable is varied at a time, in the case of high-dimensional data with multiple variables varied simultaneously, as is the case in additive manufacturing, interpreting the trends through two- or three-dimensional heat maps becomes challenging. Here, we propose a strategy of mixed visual data mining and quantitative analysis for high-dimensional metallurgical and process data using high-throughput thermodynamic calculations. Two case studies show the application of the proposed approach. The first case study investigated the effects of feedstock chemistry on the δ ferrite formation in 316L stainless steel powders used for binder jet additive manufacturing. The second case study linked Scheil–Gulliver calculations to a process model for dissimilar joining of aluminum alloys 5356 and 6111 during laser hot-wire additive manufacturing. Both cases contained thousands of calculated data points, showcasing the utility of visual data analysis through parallel coordinate plotting, Pearson correlation coefficient matrices, and scatter matrices compared to traditional process maps. These visualization techniques can be extended to many additive manufacturing problems to capture process–structure–property relationships for additively manufactured components.

36 MATERIALS SCIENCE↗
DOE OSTI2021

A Multiport DC Transformer to Enable Flexible Scalable DC as a Service

The rapid adoption of new DC loads and sources, including photovoltaic arrays, DC fast charging of electric vehicle, battery energy storage and data centers, requires a large amount of DC power conversion. A majority of new deployments also necessitate the integration of multiple DC loads and sources at one site. The traditional approach to serve these new applications relies on multiple standard power converters, each dedicated to a source or load, and results in highly customized systems, with challenging control coordination, complex protection strategies, and poor scalability. Instead, this paper proposes the concept of a multiport DC transformer (MDCT) as a modular building block for realizing a flexible, scalable DC as a Service system and address this rapidly growing need. The MDCT uses the S4T to achieve very tight control of cycle-by-cycle energy exchange between multiple ports, with high efficiency. A single multiport converter replaces 4-6 distinct converters, integrates all energy flows and manages protection. As a result, a new layered control architecture is introduced to ensure stability and scalability of the MDCT, with multiple S4T power converter building blocks connected in parallel to realize a fully modular system and reach the target power levels.

30 DIRECT ENERGY CONVERSION↗
DOE OSTI2016

Beyond union of subspaces: Subspace pursuit on Grassmann manifold for data representation

Discovering the underlying structure of a high-dimensional signal or big data has always been a challenging topic, and has become harder to tackle especially when the observations are exposed to arbitrary sparse perturbations. Here in this paper, built on the model of a union of subspaces (UoS) with sparse outliers and inspired by a basis pursuit strategy, we exploit the fundamental structure of a Grassmann manifold, and propose a new technique of pursuing the subspaces systematically by solving a non-convex optimization problem using the alternating direction method of multipliers. This problem as noted is further complicated by non-convex constraints on the Grassmann manifold, as well as the bilinearity in the penalty caused by the subspace bases and coefficients. Nevertheless, numerical experiments verify that the proposed algorithm, which provides elegant solutions to the sub-problems in each step, is able to de-couple the subspaces and pursue each of them under time-efficient parallel computation.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2014

Fast optimization and dose calculation in scanned ion beam therapy

Purpose: Particle therapy (PT) has advantages over photon irradiation on static tumors. An increased biological effectiveness and active target conformal dose shaping are strong arguments for PT. However, the sensitivity to changes of internal geometry complicates the use of PT for moving organs. In case of interfractionally moving objects adaptive radiotherapy (ART) concepts known from intensity modulated radiotherapy (IMRT) can be adopted for PT treatments. One ART strategy is to optimize a new treatment plan based on daily image data directly before a radiation fraction is delivered [treatment replanning (TRP)]. Optimizing treatment plans for PT using a scanned beam is a time consuming problem especially for particles other than protons where the biological effective dose has to be calculated. For the purpose of TRP, fast optimization and fast dose calculation have been implemented into the GSI in-house treatment planning system (TPS) TRiP98. Methods: This work reports about the outcome of a code analysis that resulted in optimization of the calculation processes as well as implementation of routines supporting parallel execution of the code. To benchmark the new features, the calculation time for therapy treatment planning has been studied. Results: Compared to the original version of the TPS, calculation times for treatment planning (optimization and dose calculation) have been improved by a factor of 10 with code optimization. The parallelization of the TPS resulted in a speedup factor of 12 and 5.5 for the original version and the code optimized version, respectively. Hence the total speedup of the new implementation of the authors' TPS yielded speedup factors up to 55. Conclusions: The improved TPS is capable of completing treatment planning for ion beam therapy of a prostate irradiation considering organs at risk in this has been overseen in the review process. Also see below 6 min.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2013

HykSort: a new variant of hypercube quicksort on distributed memory architectures. In: ICS '13 Proceedings of the 27th international ACM conference on International conference on supercomputing

In this paper, we present HykSort, an optimized comparison sort for distributed memory architectures that attains more than 2× improvement over bitonic sort and samplesort. The algorithm is based on the hypercube quicksort, but instead of a binary recursion, we perform a k-way recursion in which the pivots are selected accurately with an iterative parallel select algorithm. The single-node sort is performed using a vectorized and multithreaded merge sort. The advantages of HykSort are lower communication costs, better load balancing, and avoidance of O(p)-collective communication primitives. We also present a staged communication samplesort, which is more robust than the original samplesort for large core counts. We conduct an experimental study in which we compare hypercube sort, bitonic sort, the original samplesort, the staged samplesort, and HykSort. We report weak and strong scaling results and study the effect of the grain size. It turns out that no single algorithm performs best and a hybridization strategy is necessary. As a highlight of our study, on our largest experiment on 262,144 AMD cores of the CRAY XK7 "Titan" platform at the Oak Ridge National Laboratory we sorted 8 trillion 32-bit integer keys in 37 seconds achieving 0.9TB/s effective throughput.

Sundar, Hari↗
DOE OSTI2024

Accelerated Constrained Sparse Tensor Factorization on Massively Parallel Architectures

This study presents the first constrained sparse tensor factorization (cSTF) framework that optimizes and fully offloads computation to massively parallel GPU architectures, and the first performance characterization of cSTF on GPU architectures. In contrast to prior work on tensor factorization, where the matricized tensor times Khatri-Rao product (MTTKRP) is the primary performance bottleneck, our systematic analysis of the cSTF algorithm on GPUs reveals that adding constraints creates an additional bottleneck in the update operation for many real-world sparse tensors. While executing the update operation on the GPU brings significant speedup over its CPU counterpart, it remains a significant bottleneck. To further accelerate the update operation, we propose cuADMM, a new update algorithm that leverages algorithmic and code optimization strategies to minimize both computation and data movement on GPUs. As a result, our framework delivers significantly improved performance compared to prior state-of-the-art. On 10 real-world sparse tensors, our framework achieves geometric mean speedup of 5.1 × (max 41.59 ×) and 7.01 × (max 58.05 ×) on the NIVIDA A100 and H100 GPUs, respectively, over the state-of-the-art SPLATT library running on a 26-core Intel Ice Lake Xeon CPU.

Soh, Yongseok↗