Search NASA⌕ Search

SEARCH · Search NASA

Results for “task 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.

At least 523 records · Page 29

Photon Organ and Effective Dose Coefficients for Cranial and Caudal Irradiation Geometries - Paper 148

With the introduction of new recommendations of the ICRP in Publication 103 the methodology for determining the protection quantity effective dose has been modified. The modifications include changes in organ/tissue weighting factors as well as some changes in radiation weighting factors. Computations of protection dose coefficients now require the use of reference male and female voxel phantoms, see ICRP Publication 110. Computations of equivalent doses in organs and tissues are now performed in each phantom and the average is used to compute the effective dose. The latest set of external dose coefficients using the new reference phantoms and the latest guidance from the ICRP were performed by the ICRP DOCAL task group and reported in ICRP Publication 116, the revision of ICRP Publication 74 and ICRU Report 57. The coefficients were determined for the following irradiation geometries: anterior-posterior (AP), posterior-anterior (PA), right and left lateral (RLAT and LLAT), rotational (ROT), and isotropic (ISO). For all geometries, except the ISO geometry, parallel beams are used to compute the dose coefficients. In this paper, the methodology of ICRP Publication 116 was used to compute dose coefficients for photons for irradiations of the body with parallel beams directed upward from below the feet (caudal) and directed downward from above the head (cranial). These geometries may be encountered in the workplace for personnel standing on contaminated surfaces or volumes and from overhead sources. In this work calculations of organ and tissue kerma and absorbed doses for caudal and cranial exposures to photons ranging in energy from 10 keV to 10 GeV have been performed using the MCNP6.1 radiation transport code and the adult reference phantoms of ICRP Publication 110. Dose coefficients for caudal and cranial irradiation geometries have been calculated in accordance with ICRP Publication 103. The coefficients differ significantly from AP values for both organ and tissues of ICRP Publication 116, as would be expected. For workers in these scenarios organ and effective doses can differ significantly from AP values. For common photon energies (i.e. 100 keV to a few MeV) the AP effective dose overestimates the caudal and cranial values by about a factor of ten. (authors)

61 RADIATION PROTECTION AND DOSIMETRY↗

SU-D-206-06: Task-Specific Optimization of Scintillator Thickness for CMOS-Detector Based Cone-Beam Breast CT

Purpose: To optimize the cesium iodide (CsI:Tl) scintillator thickness in a complimentary metal-oxide semiconductor (CMOS)-based detector for use in dedicated cone-beam breast CT. Methods: The imaging task considered was the detection of a microcalcification cluster comprising six 220µm diameter calcium carbonate spheres, arranged in the form of a regular pentagon with 2 mm spacing on its sides and a central calcification, similar to that in ACR-recommended mammography accreditation phantom, at a mean glandular dose of 4.5 mGy. Generalized parallel-cascades based linear systems analysis was used to determine Fourier-domain image quality metrics in reconstructed object space, from which the detectability index inclusive of anatomical noise was determined for a non-prewhitening numerical observer. For 300 projections over 2π, magnification-associated focal-spot blur, Monte Carlo derived x-ray scatter, K-fluorescent emission and reabsorption within CsI:Tl, CsI:Tl quantum efficiency and optical blur, fiberoptic plate transmission efficiency and blur, CMOS quantum efficiency, pixel aperture function and additive noise, and filtered back-projection to isotropic 105µm voxel pitch with bilinear interpolation were modeled. Imaging geometry of a clinical prototype breast CT system, a 60 kV Cu/Al filtered x-ray spectrum from 0.3 mm focal spot incident on a 14 cm diameter semi-ellipsoidal breast were used to determine the detectability index for 300–600 µm thick (75µm increments) CsI:Tl. The CsI:Tl thickness that maximized the detectability index was considered optimal. Results: The limiting resolution (10% modulation transfer function, MTF) progressively decreased with increasing CsI:Tl thickness. The zero-frequency detective quantum efficiency, DQE(0), in projection space increased with increasing CsI:Tl thickness. The maximum detectability index was achieved with 525µm thick CsI:Tl scintillator. Reduced MTF at mid-to-high frequencies for 600µm thick CsI:Tl lowered the detectability index than 525µm CsI:Tl. Conclusion: For the x-ray spectrum and imaging conditions considered, a 525µm thick CsI:Tl scintillator integrated with the CMOS detector is optimal for detecting microcalcification cluster. Funding support: Supported in part by NIH R01 CA195512. The contents are solely the responsibility of the authors and do not reflect the official views of the NIH or the NCI. Disclosures: SV, GV and AK - Research collaboration, Koning Corp., West Henrietta, NY.

60 APPLIED LIFE SCIENCES↗

Methods and Tools to allow molecular flow simulations to be coupled to higher level continuum descriptions of flows in porous/fractured media and aerosol/dust dynamics

The purpose of this project was to develop methods and tools that will aid in safety evaluation of nuclear fuels and licensing of nuclear reactors relating to accidents.The objectives were to develop more detailed and faster computations of fission product transport and aerosol evolution as they generally relate to nuclear fuel and/or nuclear reactor accidents. The two tasks in the project related to molecular transport in nuclear fuel and aerosol transport in reactor vessel and containment. For both the tasks, explorations of coupling of Direct Simulation Monte Carlo with Navier-Stokes solvers or the Sectional method were not successful. However, Mesh free methods for the Direct Simulation Monte Carlo method were successfully explored.These explorations permit applications to porous and fractured media, and arbitrary geometries.The computations were carried out in Mathematica and are fully parallelized. The project has resulted in new computational tools (algorithms and programs) that will improve the fidelity of computations to actual physics, chemistry and transport of fission products in the nuclear fuel and aerosol in reactor primary and secondary containments.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Asynchronous-many-task systems: Challenges and opportunities - Scaling an AMR astrophysics code on exascale machines using Kokkos and HPX

Dynamic and adaptive mesh refinement is pivotal in high-resolution, multi-physics, multi-model simulations, necessitating precise physics resolution in localized areas across expansive domains. Today’s supercomputers’ extreme heterogeneity presents a significant challenge for dynamically adaptive codes, highlighting the importance of achieving performance portability at scale. Our research focuses on astrophysical simulations, particularly stellar mergers, to elucidate early universe dynamics. Here, we present Octo-Tiger, leveraging Kokkos, HPX, and SIMD for portable performance at scale in complex, massively parallel adaptive multi-physics simulations. Octo-Tiger supports diverse processors, accelerators, and network backends. Experiments demonstrate exceptional scalability across several heterogeneous supercomputers including Perlmutter, Frontier, and Fugaku, encompassing major GPU architectures and x86, ARM, and RISC-V CPUs. Parallel efficiency of 47.59% (110,080 cores and 6880 hybrid A100 GPUs) on a full-system run on Perlmutter (26% HPCG peak performance) and 51.37% (using 32,768 cores and 2048 MI250X) on Frontier are achieved.

97 MATHEMATICS AND COMPUTING↗

An Adaptive-Mesh-Refinement Based Computational Tool for Simulating Catalysis at Mesoscale

In this work, we present a computational tool for mesoscale applications using open-source exascale- computing compatible adaptive-mesh-refinement (AMR) library, AMReX [2]. AMReX is software library that enables development of application solvers with block-structured Cartesian AMR. Our tool has capabilities to include realistic geometry representation, chemical species transport, reactions and thermodynamics that are critical for capturing mesoscale physics. A significant achievement is the ability of our solver to automatically import electron microscopy data in the form of a stereolithography (STL) or pixelated file format (mrc, tiff) without undergoing the tedious task of unstructured mesh generation. This feature allows for rapid simulation of catalyst particles with complex morphologies using an immersed-boundary formulation. The use of AMR allows for higher resolutions at catalyst surface interfaces, which in turn provides an accurate description of surface reactions and transport. Our solver uses a hybrid distributed and shared memory parallelism (OpenMP/GPU-based) with which strong scaling up to 10,000 processors for realistic catalyst particle simulations have been demonstrated.

BIOMASS FUELS,MATHEMATICS AND COMPUTING↗

PRIMA-X - Performance Retargeting of Instrumentation, Measurement, and Analysis Technologies for Exascale Computing

The PRIMA-X (Performance Retargeting of Instrumentation, Measurement, and Analysis Technologies for Exascale Computing) project is the successor of the DOE PRIMA (Performance Refactoring of Instrumentation, Measurement, and Analysis Technologies for Petascale Computing) project, which addressed the challenge of creating a core measurement infrastructure that would serve as a common platform for both integrating leading parallel performance systems (notably TAU and Scalasca) and developing next-generation scalable performance tools. The PRIMA-X project shifts the focus away from refactorization of robust performance tools towards a re-targeting of the parallel performance measurement and analysis architecture for extreme scales. The massive concurrency, asynchronous execution dynamics, hardware heterogeneity, and multi-objective prerequisites (performance, power, resilience) that identify exascale systems introduce fundamental constraints on the ability to carry forward existing performance methodologies. In particular, there must be a deemphasis of per-thread observation techniques to significantly reduce the otherwise unsustainable flood of redundant performance data. Instead, it will be necessary to assimilate multi-level resource observations into macroscopic performance views, from which resilient performance metrics can be attributed to the computational features of the application. This requires a scalable framework for node-level and system-wide monitoring and runtime analyses of dynamic performance information. Also, the interest in optimizing parallelism parameters with respect to performance and energy drives the integration of tool capabilities in the exascale environment further. Initially, PRIMA-X was a collaborative project between the University of Oregon (lead institution) and the German Research School for Simulation Sciences (GRS). Because Prof. Wolf, the PI at GRS, accepted a position as full professor at Technische Universitaet Darmstadt (TU Darmstadt) starting February 1st, 2015, the work of GRS was continued at TU Darmstadt. This report covers the work at TU Darmstadt after the transition. The first main accomplishment of TU Darmstadt is the development of different techniques to aggregate performance data on the level of threads. TU Darmstadt evaluated different schemes designed at GRS during the first phase of this project and integrated them into Score-P, a widely used performance measurement framework for HPC applications. The second main accomplishment is a substantial increase of Score-P’s scalability, achieved by improving its internal representation of the underlying system. Third, we extended our performance-modeling tool Extra-P to automatically create models with more than one input parameter (e.g., process count and input size), which was previously impossible. These models allow the performance behavior of an application to be extrapolated to larger machines. Specific applications of the new multi-parameter capability realized in this project include modeling the isoefficiency function of task-based applications and a novel co-design methodology. Further extensions of Extra-P comprise online modeling functionality, automatic refinement of the model search space, and segmented performance modeling. Further contributions include new locality metrics for both shared- and distributed-memory applications, a new way of measuring the impact of inter-application inference on the execution time of an application, and finally an automatic method that selects appropriate OpenMP constructs for the parallelization of sequential code and classifies the affected variables based on their sharing semantics.

97 MATHEMATICS AND COMPUTING↗

Advanced Laboratory and Field Arrays: Evaluating Sampling Techniques for MHK Biological Monitoring (Task 6)

The overall goal of the task was to identify cost effective biological sampling techniques for MHK environmental monitoring. Protected, demersal, and pelagic fish and selected nektonic invertebrates were surveyed using capture and remote sensing techniques at the PacWave sites. The performance of capture and remote sensing monitoring techniques were evaluated and generic nekton monitoring indices were developed for MHK technologies and sites. In parallel to data collections at the PacWave site, the ability of regression models to characterize, detect, and predict change in acoustic data was evaluated using acoustic data collected in Admiralty Inlet, WA, during a biological monitoring study of the proposed SnoPud tidal turbine project site. Expected outcomes of these efforts included an evaluation of instrumentation and techniques used to monitor biological variability; identification of data streams that can be used to detect and quantify change; and sampling requirements to ensure detection of change in monitored variables.

13 HYDRO ENERGY↗

Evaluating Corrosion Effects on the Stainless Steel Components of the SAVY-4000/Hagan Nuclear Material Storage Containers: FY2022 Update

This report summarizes the work completed and underway for the Corrosion Working Group (CWG) for FY2022 in support of evaluating a greater than a 15 year design life for the SAVY4000. The CWG meets on an as needed basis to review new work, results or information regarding the corrosion of SAVY-4000 and Hagan nuclear material storage containers. The goal when analyzing corrosion is to develop a design life appropriate for the containers and to identify when a container should be removed from service. There are many parallel efforts associated with identifying and extending the service life of nuclear material storage containers. This report includes information from research and development efforts to address the lifetime. The CWG is closely related to surveillance activities and continues to play a role in identification of issues and paths forward for analysis of surveillance containers. This work summarizes the completed tasks for FY22 of the CWG while keeping in mind that some efforts continue into future years. Brief summaries of identified new avenues of work have been identified for FY23.

36 MATERIALS SCIENCE↗

Requirements Description of GAMSOR and GAMSRC

This report presents the modeling and simulation capabilities of Argonne National Laboratory’s GAMSOR and GAMSRC code that is used in present reactor analysis activities. These capabilities will then be used to establish the set of verification tasks necessary to verify GAMSOR for use on commercial projects. A similar approach was taken for the PERSENT, REBUS and DIF3D software packages. The GAMSOR program is a modified version of DIF3D used to obtain the coupled neutron gamma flux distributions for a given reactor system. GAMSOR was first developed in the 1980s and its work has progressed in parallel with DIF3D throughout its various development phases. While GAMSOR is compatible with DIF3D-FD and DIF3D-Nodal, DIF3D-VARIANT is most commonly used today and will be the focus of the present verification effort. As covered in [1], GAMSOR was built to provide more accurate heating distributions by explicitly solving for both the gamma and neutron heating distributions. Without this consideration, gamma heat is normally assumed to be deposited locally at the site of gamma production. This is acceptable for modeling some phenomena because the total energy production is preserved, but it can lead to inaccurate approximations of fluence-dependent behaviors such as fuel performance, swelling, cladding embrittlement, radiation induced creep, and power peaking. The outputs from GAMSOR are contained in multiple intermediate and final output files: FIXSRC, VARSRC, GTFLUX, GHFLUX, PWDINT, NPDINT, and GPDINT. DIF3D normally produces a PWDINT file containing power density results, with the assumption that gamma rays are absorbed where they are emitted (i.e., energy deposition is local with no modeled transport). With GAMSOR, neutron power density (NPDINT) and gamma power density (GPDINT) are both produced, with their summed power densities replacing the results in PWDINT for total heating (i.e., recoverable) power density.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Encoder–decoder neural network for solving the nonlinear Fokker–Planck–Landau collision operator in XGC

An encoder–decoder neural network has been used to examine the possibility for acceleration of a partial integro-differential equation, the Fokker–Planck–Landau collision operator. This is part of the governing equation in the massively parallel particle-in-cell code XGC, which is used to study turbulence in fusion energy devices. The neural network emphasizes physics-inspired learning, where it is taught to respect physical conservation constraints of the collision operator by including them in the training loss, along with the ℓ 2 loss. In particular, network architectures used for the computer vision task of semantic segmentation have been used for training. A penalization method is used to enforce the ‘soft’ constraints of the system and integrate error in the conservation properties into the loss function. During training, quantities representing the particle density, momentum and energy for all species of the system are calculated at each configuration vertex, mirroring the procedure in XGC. This simple training has produced a median relative loss, across configuration space, of the order of 10 –4 , which is low enough if the error is of random nature, but not if it is of drift nature in time steps. The run time for the current Picard iterative solver of the operator is O(n 2 ), where n is the number of plasma species. As the XGC1 code begins to attack problems including a larger number of species, the collision operator will become expensive computationally, making the neural network solver even more important, especially since its training only scales as O(n). Here, a wide enough range of collisionality has been considered in the training data to ensure the full domain of collision physics is captured. An advanced technique to decrease the losses further will be subject of a subsequent report. Eventual work will include expansion of the network to include multiple plasma species.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Rigorous Data Fusion for Computationally Expensive Simulations

This manuscript comprises the final report for the 1-year, FY19 LDRD project "Rigorous Data Fusion for Computationally Expensive Simulations," wherein an alternative approach to Bayesian calibration was developed based a new sampling technique called VoroSpokes. Vorospokes is a novel quadrature and sampling framework defined with respect to Voronoi tessellations of bounded domains in $R^d$ developed within this project. In this work, we first establish local quadrature and sampling results on convex polytopes using randomly directed rays, or spokes, to approximate the quantities of interest for a specified target function. A theoretical justification for both procedures is provided along with empirical results demonstrating the unbiased convergence in the resulting estimates/samples. The local quadrature and sampling procedures are then extended to global procedures defined on more general domains by applying the local results to the cells of a Voronoi tessellation covering the domain in consideration. We then demonstrate how the proposed global sampling procedure can be used to define a natural framework for adaptively constructing Voronoi Piecewise Surrogate (VPS) approximations based on local error estimates. Finally, we show that the adaptive VPS procedure can be used to form a surrogate model approximation to a specified, potentially unnormalized, density function, and that the global sampling procedure can be used to efficiently draw independent samples from the surrogate density in parallel. The performance of the resulting VoroSpokes sampling framework is assessed on a collection of Bayesian inference problems and is shown to provide highly accurate posterior predictions which align with the results obtained using traditional methods such as Gibbs sampling and random-walk Markov Chain Monte Carlo (MCMC). Importantly, the proposed framework provides a foundation for performing Bayesian inference tasks which is entirely independent from the theory of Markov chains.

97 MATHEMATICS AND COMPUTING↗

Formation of Imploding Plasma Liners for HEDP and MIF Application

Plasma jets with high density and velocity have a number of important applications in fusion energy and elsewhere, including plasma refueling, disruption mitigation in tokamaks, magnetized target fusion, injection of momentum into centrifugally confined mirrors, plasma thrusters, and high energy density plasmas (HEDP). In Magneto-Inertial Fusion (MIF), for example, an imploding material liner is used to compress a magnetized plasma to fusion conditions and to confine the resulting burning plasma inertially to obtain the necessary energy gain. The imploding shell may be solid, liquid, gaseous, or a combination of these states. The presence of the magnetic field in the target plasma suppresses thermal transport to the plasma shell, thus lowering the imploding power needed to compress the target to fusion conditions. This allows the required imploding momentum flux to be generated electromagnetically using off-the-shelf pulsed power technology. Practical schemes for standoff delivery of the imploding momentum flux are required and are open topics for research. One approach for accomplishing this, called plasma jet driven magneto-inertial fusion (PJMIF), uses a spherical array of pulsed plasma guns to create a spherically imploding shell of very high velocity, high momentum flux plasma. This approach requires development of plasma jet accelerators capable of achieving velocities of 50-200 km/s with very precise timing and density profiles, and with high total mass and density. Low-Z plasma jets would require the higher velocities, whereas very dense high-Z plasma shells could achieve the goal at velocities of only 50-100 km/s. In this report, we describe our work to develop the pulsed plasma gun technology needed for an experimental scientific exploration of the PJMIF concept, and also for the other applications mentioned earlier. The initial goal of a few hundred of hydrogen at 200 km/s was eventually replaced with accelerating 8000 μg of argon or xenon to 50 km/s for the Plasma Liner Experiment (PLX) at Los Alamos National Laboratory (LANL). Initial work used existing computational and analytical tools to develop and refine a specific plasma gun concept having a novel tapered coaxial electromagnetic accelerator contour with an array of symmetric ablative plasma injectors. The profile is designed to suppress the main barrier to success in coaxial guns, namely the blow-by instability in which the arc slips past and outruns the bulk of the plasma mass. Efforts to begin developing a set of annular non-ablative plasma injectors for the coaxial gun, in order to accelerate pure gases, resulted in development of linear parallel-plate MiniRailguns that turned out to work well as plasma guns in their own right and we subsequently chose them for an initial plasma liner experiment on the PLX facility at LANL. This choice was mainly driven by cost and schedule for that particular experiment, while longer term goals still projected use of coaxial guns for reactor-relevant applications for reasons of better symmetry, lower impurities, more compact plasma jet formation, and higher gun efficiency. Our efforts have focused mainly on 1) developing various plasma injection systems for both coax and linear railguns and ensuring they work reliably with the accelerator section, 2) developing a suite of plasma and gun diagnostics, 3) performing computational modeling to design and refine the plasma guns, 4) establishing a research facility dedicated to plasma gun development, and finally, 5) developing plasma guns and associated pulse power systems capable of achieving these goals and installing and testing the first two gun sets on the PLX facility at LANL. During the second funding cycle for this program, HyperV joined in a collaborative effort with LANL, the University of Alabama at Huntsville, and the University of New Mexico to perform a plasma liner experiment (PLX) to investigate the physics and technology of forming spherically imploding plasma liners. HyperV’s tasks focused on developing the plasma guns and associated pulse power systems required for the 30 gun experiment at LANL. Unfortunately, funding for the entire PLX collaborative project was terminated after only two years of the four year project due to program funding realignments which necessitated recompeting the project in midstream. Despite the loss of funding, HyperV installed two Mark1 guns and pulsed power systems on PLX, and jet characterization and merging experiments were subsequently successfully performed at LANL by the PLX Team. In parallel with those PLX experiments, HyperV continued its efforts to develop a plasma gun capable of meeting the PLX goal of 8 mg of argon at 50 km/s. HyperV was ultimately successful in this effort, demonstrating 10.8 mg at 52.8 km/s and 7.5 mg at 62.4 km/s with the Mark2 MiniRailgun.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Acoustophoretic Additive Manufacturing for Scalable 3D Battery Electrodes

This project focused on investigating two acoustic-based processing methods: a nozzle-based printhead and a chamber that map to two different battery electrode architectures: (1) a line-pattern electrode and (2) a grid-pattern electrode. These two parallel manufacturing and electrode geometry explorations were proposed for the project to understand the process space of acoustic-based manufacturing methods to fabricate patterned battery electrodes. This two-path exploration also allowed us to de-risk the overall project and not rely on a single process for creating patterned electrodes. This project consisted of six high-level tasks aimed at transitioning the concept of acoustic focusing for battery electrodes from a technology readiness level (TRL) of 1 to 3 by project conclusion. Overall, we believe we have developed a practical and high-impact processing method that is chemistry agnostic and suitable for large-area fabrication of both 3D LIBs and other functional material systems where structuring on the scale of tens of microns has the potential to break conventional bulk material trade-offs in performance. In the case of batteries for electric vehicles, structuring 3D LIBs with our acoustic process breaks traditional energy and power trade-offs observed with conventional flat battery packs.

25 ENERGY STORAGE↗

HEP Software Foundation Community White Paper Working Group - Data Processing Frameworks

Data processing frameworks are an essential part of HEP experiments' software stacks. Frameworks provide a means by which code developers can undertake the essential tasks of physics data processing, accessing relevant inputs and storing their outputs, in a coherent way without needing to know the details of other domains. Frameworks provide essential core services for developers and help deliver a configurable working application to the experiments' production systems. Modern HEP processing frameworks are in the process of adapting to a new computing landscape dominated by parallel processing and heterogeneity, which pose many questions regarding enhanced functionality and scaling that must be faced without compromising the maintainability of the code. In this paper we identify a program of work that can help further clarify the key concepts of frameworks for HEP and then spawn R&D activities that can focus the community's efforts in the most efficient manner to address the challenges of the upcoming experimental program.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Efficient Treatment of Large Active Spaces through Multi-GPU Parallel Implementation of Direct Configuration Interaction

In this study, we have extended our graphical processing unit (GPU)-accelerated direct configuration interaction program to multiple devices, reducing iteration times for configuration spaces of 165 million determinants to only 3 s using NVIDIA P100 GPUs. Similar improvements in the one- and two-particle reduced density matrix formation allow for fast analytical energy gradients and electronic properties. Our parallel algorithm enables the calculation of arbitrarily large configuration spaces (limited only by available system memory), with iteration times of 13 min for an active space of 18 electrons in 18 orbitals (2.4 billion determinants) using six consumer grade NVIDIA 1080Ti GPUs. These advances enable routine molecular dynamics simulations, geometry optimizations, and absorption spectrum calculations for molecules with large configuration spaces, a task that has heretofore required massive computational effort. In this work, we demonstrate the utility of our program by generating the absorption spectrum for diphenyl acetylene at the floating occupation molecular orbital complete active space configuration interaction level of theory. Lastly, several active spaces were investigated to assess the dependence of spectral features on orbital space dimension.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

A survey on checkpointing strategies: Should we always checkpoint à la Young/Daly?

The Young/Daly formula provides an approximation of the optimal checkpointing period for a parallel application executing on a supercomputing platform. It was originally designed to handle fail-stop errors for preemptible tightly-coupled applications, but has been extended to other application and resilience frameworks. Here, we provide some background and survey various scenarios to assess the usefulness and limitations of the formula, both for preemptible applications and workflow applications represented as a graph of tasks. We also discuss scenarios with uncertainties, and extend the study to silent errors. We exhibit cases where the optimal period is of a different order than that dictated by the Young/Daly formula, and finally we explain how checkpointing can be further combined with replication.

97 MATHEMATICS AND COMPUTING↗

Investigation robot on cables - 15250

The Decontamination and Decommissioning (D and D) Operations take place in hostile environments to the human intervention. In this context, robotic systems are a solution to achieve these operations while minimizing human intervention. Among these operations, the monitoring of lengthwise plant has to be performed in places where the radioactive contaminations make hostile environments. These monitoring operations need to measure several physics parameters in several locations in a precise and safe manner. These repetitive tasks can be made by robots that are simple, small sized and easily adaptable. In this scope, a robotic system on cables has been designed to control and monitor the inside of lengthwise nuclear facilities. In lengthwise installations, the using of cables allows to increase the covered zone of the robots which is an essential quality in such places. Unlike cable-driven parallel robots used in workspaces of 2D or 3D controlled by redundant tensioning motors, the current robot moves in a 1D workspace. Thus a tractors cable controlled by one tensioning motor is used to drive the robot on carrier cables. In this paper, we will describe the first results of this work by bringing out the innovation of our system. Then, we will discuss the perspective to upgrade the actual system whose applications could be perfectly imagined outside the nuclear D and D industry. (authors)

42 ENGINEERING↗

FPDeep: Scalable Acceleration of CNN Training on Deeply-Pipelined FPGA Clusters

In this paper, we propose a framework called FPDeep, which uses a hybrid of model and layer paral- lelism to configure distributed reconfigurable clusters to train DNNs. This approach has numerous benefits. First, the design does not suffer from batch size growth. Second, novel workload and weight partitioning leads to balanced loads of both among nodes. And third, the entire system is fine-grained pipeline. This leads to high parallelism and utilization and also minimizes the time features need to be cached while waiting for back-propagation.

Wang, Tianqi↗