Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Data Intensive Computing”

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 1,171 records · Page 65

Numerical Modeling of Laser Heating and Evaporation of a Single Droplet

Laser technology is being widely studied for controlled energy deposition for a range of applications, including flow control, ignition, combustion, and diagnostics. The absorption and scattering of laser radiation by liquid droplets in aerosols affects propagation of the laser beam in the atmosphere, while the ignition and combustion characteristics in combustion chambers are influenced by the evaporation rate of the sprayed fuel. In this work, we present a mathematical model built on OpenFOAM for laser heating and evaporation of a single droplet in the diffusion-dominated regime taking into account absorption of the laser radiation, evaporation process, and vapor flow dynamics. The developed solver is validated against available experimental and numerical data for heating and evaporation of ethanol and water droplets. The two main regimes—continuous and pulsed laser heating—are explored. For continuous laser heating, the peak temperature is higher for larger droplets. For pulsed laser heating, when the peak irradiance is close to transition to the boiling regime, the temporal dynamics of the droplet temperature does not depend on the droplet size. With the empirical normalization of time, the dynamics of the droplet shrinkage and cooling are found to be independent of droplet sizes and peak laser intensities.

42 ENGINEERING↗

Reconstruction of Six-Dimensional Phase Space

A phase space is a mathematical representation of all possible physical states of a system. Particle beams at Fermilab exist within a six-dimensional (6D) phase space defined by three positional components, (x, y, z) and three momentum components, (px, py, pz). To reconstruct this space implies taking measurement data from detectors and mapping out particle behavior using computational methods. The beam detectors, however, are only able to detect spatial distribution among the events of the beam, therefore being limited to positional data. Also, due to the vast number of events in a particle beam, it is extremely difficult to analyze and differentiate every single one’s behavior. However, with Machine Learning (ML), which can distinguish between patterns and map out particle behavior more efficiently. We first used the particle beam software, G4beamline, to simulate a 10,000-event muon beam, adjusting parameters such as initial momentum magnitude (p¬0) and virtual detector position. Using ten virtual detectors, we analyzed p0 values such that minimum 9,990 events were analyzed by every detector. We then input the data from these beam simulations to a C++ program, that randomly selects 100 events, and creates a 2D histogram based on spatial distribution, detector position, and event intensity. This process is repeated 100 times to create 100 histograms per p0 value. These images were then input to a modified ResNet18 Convolutional Neural Network (CNN) for training, and to predict p0 from some unseen set of histograms. The model was accurate when trained on momentum increments of 5 MeV/c and provided with denser training samples around highly variable test values. These results displayed machine learning being able to accurately predict p0 from being trained on different particle behaviors.

Shirlee, Jermain [Fermilab]↗

SU-G-BRA-04: Simulation of Errors in Maximal Intensity Projection (MIP)-Based Lung Tumor Internal Target Volumes (ITV) Using Real-Time 2D MRI and Deformable Image Registration Based Lung Tumor Tracking

Purpose: To evaluate the effect of inter- and intra-fractional tumor motion on the error in four-dimensional computed tomography (4DCT) maximal intensity projection (MIP)–based lung tumor internal target volumes (ITV), using deformable image registration of real-time 2D-sagital cine-mode MRI acquired during lung SBRT treatments. Methods: Five lung tumor patients underwent free breathing SBRT treatment on the ViewRay, with dose prescribed to PTV (4DCT MIP-based ITV+3–6mm margin). Sagittal slice cine-MR images (3.5×3.5mm pixels) were acquired through the center of the tumor at 4 frames per second throughout the treatments (3–4 fractions of 21–32 minutes duration). Tumor GTVs were contoured on the first frame of the cine and tracked throughout the treatment using off-line optical-flow based deformable registration implemented on a GPU cluster. Pseudo-4DCT MIP-based ITVs were generated from MIPs of the deformed GTV contours limited to short segments of image data. All possible pseudo-4DCT MIP-based ITV volumes were generated with 1s resolution and compared to the ITV volume of the entire treatment course. Varying pseudo-4DCT durations from 10-50s were analyzed. Results: Tumors were covered in their entirety by PTV in the patients analysed here. However, pseudo-4DCT based ITV volumes were observed that were as small as 29% of the entire treatment-ITV, depending on breathing irregularity and the duration of pseudo-4DCT. With an increase in duration of pseudo-4DCT from 10–50s the minimum volume acquired from 95% of all pseudo-4DCTs increased from 62%–81% of the treatment ITV. Conclusion: A 4DCT MIP-based ITV offers a ‘snap-shot’ of breathing motion for the brief period of time the tumor is imaged on a specific day. Real time MRI over prolonged periods of time and over multiple treatment fractions shows that the accuracy of this snap-shot varies according to inter- and intra-fractional tumor motion. Further work is required to investigate the dosimetric effect of these results.

60 APPLIED LIFE SCIENCES↗

Sirepo - Warp

Sirepo is an open source framework for cloud computing. The graphical user interface (GUI) for Sirepo, also known as the client, executes in any HTML5 compliant web browser on any computing platform, including tablets. The client is built in JavaScript, making use of the following open source libraries: Bootstrap, which is fundamental for cross-platform web applications; AngularJS, which provides a model–view–controller (MVC) architecture and GUI components; and D3.js, which provides interactive plots and data-driven transformations. The Sirepo server is built on the following Python technologies: Flask, which is a lightweight framework for web development; Jin-ja, which is a secure and widely used templating language; and Werkzeug, a utility library that is compliant with the WSGI standard. We use Nginx as the HTTP server and proxy, which provides a scalable event-driven architecture. The physics codes supported by Sirepo execute inside a Docker container. One of the codes supported by Sirepo is Warp. Warp is a particle-in-cell (PIC) code de-signed to simulate high-intensity charged particle beams and plasmas in both the electrostatic and electromagnetic regimes, with a wide variety of integrated physics models and diagnostics. At pre-sent, Sirepo supports a small subset of Warp’s capabilities. Warp is open source and is part of the Berkeley Lab Accelerator Simulation Toolkit.

Nagler, Robert↗

Sirepo - Warp

Sirepo is an open source framework for cloud computing. The graphical user interface (GUI) for Sirepo, also known as the client, executes in any HTML5 compliant web browser on any computing platform, including tablets. The client is built in JavaScript, making use of the following open source libraries: Bootstrap, which is fundamental for cross-platform web applications; AngularJS, which provides a model–view–controller (MVC) architecture and GUI components; and D3.js, which provides interactive plots and data-driven transformations. The Sirepo server is built on the following Python technologies: Flask, which is a lightweight framework for web development; Jin-ja, which is a secure and widely used templating language; and Werkzeug, a utility library that is compliant with the WSGI standard. We use Nginx as the HTTP server and proxy, which provides a scalable event-driven architecture. The physics codes supported by Sirepo execute inside a Docker container. One of the codes supported by Sirepo is Warp. Warp is a particle-in-cell (PIC) code de-signed to simulate high-intensity charged particle beams and plasmas in both the electrostatic and electromagnetic regimes, with a wide variety of integrated physics models and diagnostics. At pre-sent, Sirepo supports a small subset of Warp’s capabilities. Warp is open source and is part of the Berkeley Lab Accelerator Simulation Toolkit.

Nagler, Robert↗

Visibility Histograms and Visibility-Driven Transfer Functions

Direct volume rendering is an important tool for visualizing complex data sets. However, in the process of generating 2D images from 3D data, information is lost in the form of attenuation and occlusion. The lack of a feedback mechanism to quantify the loss of information in the rendering process makes the design of good transfer functions a difficult and time consuming task. In this paper, we present the general notion of visibility histograms, which are multidimensional graphical representations of the distribution of visibility in a volume-rendered image. In this paper, we explore the 1D and 2D transfer functions that result from intensity values and gradient magnitude. With the help of these histograms, users can manage a complex set of transfer function parameters that maximize the visibility of the intervals of interest and provide high quality images of volume data. We present a semiautomated method for generating transfer functions, which progressively explores the transfer function space toward the goal of maximizing visibility of important structures. Our methodology can be easily deployed in most visualization systems and can be used together with traditional 1D and 2D opacity transfer functions based on scalar values, as well as with other more sophisticated rendering algorithms.

42 ENGINEERING↗

A multifractal approach to space-filling recovery for PET quantification

Purpose: A new image-based methodology is developed for estimating the apparent space-filling properties of an object of interest in PET imaging without need for a robust segmentation step and used to recover accurate estimates of total lesion activity (TLA). Methods: A multifractal approach and the fractal dimension are proposed to recover the apparent space-filling index of a lesion (tumor volume, TV) embedded in nonzero background. A practical implementation is proposed, and the index is subsequently used with mean standardized uptake value (SUV {sub mean}) to correct TLA estimates obtained from approximate lesion contours. The methodology is illustrated on fractal and synthetic objects contaminated by partial volume effects (PVEs), validated on realistic {sup 18}F-fluorodeoxyglucose PET simulations and tested for its robustness using a clinical {sup 18}F-fluorothymidine PET test–retest dataset. Results: TLA estimates were stable for a range of resolutions typical in PET oncology (4–6 mm). By contrast, the space-filling index and intensity estimates were resolution dependent. TLA was generally recovered within 15% of ground truth on postfiltered PET images affected by PVEs. Volumes were recovered within 15% variability in the repeatability study. Results indicated that TLA is a more robust index than other traditional metrics such as SUV {sub mean} or TV measurements across imaging protocols. Conclusions: The fractal procedure reported here is proposed as a simple and effective computational alternative to existing methodologies which require the incorporation of image preprocessing steps (i.e., partial volume correction and automatic segmentation) prior to quantification.

60 APPLIED LIFE SCIENCES↗

Hippocampal Dosimetry Predicts Neurocognitive Function Impairment After Fractionated Stereotactic Radiotherapy for Benign or Low-Grade Adult Brain Tumors

Purpose: To prospectively evaluate the association between hippocampal dose and long-term neurocognitive function (NCF) impairment for benign or low-grade adult brain tumors treated with fractionated stereotactic radiotherapy (FSRT). Methods and Materials: Adult patients with benign or low-grade adult brain tumors were treated with FSRT per institutional practice. No attempt was made to spare the hippocampus. NCF testing was conducted at baseline and 18 months follow-up, on a prospective clinical trial. Regression-based standardized z scores were calculated by using similar healthy control individuals evaluated at the same test-retest interval. NCF impairment was defined as a z score {<=}-1.5. After delineation of the bilateral hippocampi according to the Radiation Therapy Oncology Group contouring atlas, dose-volume histograms were generated for the left and right hippocampi and for the composite pair. Biologically equivalent doses in 2-Gy fractions (EQD{sub 2}) assuming an {alpha}/{beta} ratio of 2 Gy were computed. Fisher's exact test and binary logistic regression were used for univariate and multivariate analyses, respectively. Dose-response data were fit to a nonlinear model. Results: Of 29 patients enrolled in this trial, 18 completed both baseline and 18-month NCF testing. An EQD{sub 2} to 40% of the bilateral hippocampi >7.3 Gy was associated with impairment in Wechsler Memory Scale-III Word List (WMS-WL) delayed recall (odds ratio [OR] 19.3; p = 0.043). The association between WMS-WL delayed recall and EQD{sub 2} to 100% of the bilateral hippocampi >0.0 Gy trended to significance (OR 14.8; p = 0.068). Conclusion: EQD{sub 2} to 40% of the bilateral hippocampi greater than 7.3 Gy is associated with long-term impairment in list-learning delayed recall after FSRT for benign or low-grade adult brain tumors. Given that modern intensity-modulated radiotherapy techniques can reduce the dose to the bilateral hippocampi below this dosimetric threshold, patients should be enrolled in ongoing prospective trials of hippocampal sparing during cranial irradiation to confirm these preliminary results.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Hippocampal Dosimetry Predicts Neurocognitive Function Impairment After Fractionated Stereotactic Radiotherapy for Benign or Low-Grade Adult Brain Tumors

Purpose: To prospectively evaluate the association between hippocampal dose and long-term neurocognitive function (NCF) impairment for benign or low-grade adult brain tumors treated with fractionated stereotactic radiotherapy (FSRT). Methods and Materials: Adult patients with benign or low-grade adult brain tumors were treated with FSRT per institutional practice. No attempt was made to spare the hippocampus. NCF testing was conducted at baseline and 18 months follow-up, on a prospective clinical trial. Regression-based standardized z scores were calculated by using similar healthy control individuals evaluated at the same test-retest interval. NCF impairment was defined as a z score {<=}-1.5. After delineation of the bilateral hippocampi according to the Radiation Therapy Oncology Group contouring atlas, dose-volume histograms were generated for the left and right hippocampi and for the composite pair. Biologically equivalent doses in 2-Gy fractions (EQD{sub 2}) assuming an {alpha}/{beta} ratio of 2 Gy were computed. Fisher's exact test and binary logistic regression were used for univariate and multivariate analyses, respectively. Dose-response data were fit to a nonlinear model. Results: Of 29 patients enrolled in this trial, 18 completed both baseline and 18-month NCF testing. An EQD{sub 2} to 40% of the bilateral hippocampi >7.3 Gy was associated with impairment in Wechsler Memory Scale-III Word List (WMS-WL) delayed recall (odds ratio [OR] 19.3; p = 0.043). The association between WMS-WL delayed recall and EQD{sub 2} to 100% of the bilateral hippocampi >0.0 Gy trended to significance (OR 14.8; p = 0.068). Conclusion: EQD{sub 2} to 40% of the bilateral hippocampi greater than 7.3 Gy is associated with long-term impairment in list-learning delayed recall after FSRT for benign or low-grade adult brain tumors. Given that modern intensity-modulated radiotherapy techniques can reduce the dose to the bilateral hippocampi below this dosimetric threshold, patients should be enrolled in ongoing prospective trials of hippocampal sparing during cranial irradiation to confirm these preliminary results.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

IS ETA CARINAE A FAST ROTATOR, AND HOW MUCH DOES THE COMPANION INFLUENCE THE INNER WIND STRUCTURE?

We analyze interferometric measurements of the luminous blue variable Eta Carinae with the goal of constraining the rotational velocity of the primary star and probing the influence of the companion. Using two-dimensional radiative transfer models of latitude-dependent stellar winds, we find that prolate-wind models with a ratio of the rotational velocity (v{sub rot}) to the critical velocity (v{sub crit}) of W = 0.77-0.92, inclination angle of i = 60{sup 0}-90{sup 0}, and position angle (P.A.) =108{sup 0}-142{sup 0} reproduce simultaneously K-band continuum visibilities from VLTI/VINCI and closure phase measurements from VLTI/AMBER. Interestingly, oblate models with W = 0.73-0.90 and i = 80{sup 0}-90{sup 0} produce similar fits to the interferometric data, but require P.A. =210{sup 0}-230{sup 0}. Therefore, both prolate and oblate models suggest that the rotation axis of the primary star is not aligned with the Homunculus polar axis. We also compute radiative transfer models of the primary star allowing for the presence of a cavity and dense wind-wind interaction region created by the companion star. We find that the wind-wind interaction has a significant effect on the K-band image mainly via free-free emission from the compressed walls and, for reasonable model parameters, can reproduce the VLTI/VINCI visibilities taken at {phi}{sub vb03} = 0.92-0.93. We conclude that the density structure of the primary wind can be sufficiently disturbed by the companion, thus mimicking the effects of fast rotation in the interferometric observables. Therefore, fast rotation may not be the only explanation for the interferometric observations. Intense temporal monitoring and three-dimensional modeling are needed to resolve these issues.

79 ASTRONOMY AND ASTROPHYSICS↗

Universal parameters of bulk-solvent masks

The bulk solvent is a major component of biomacromolecular crystals that contributes significantly to the observed diffraction intensities. Accurate modelling of the bulk solvent has been recognized as important for many crystallographic calculations. Owing to its simplicity and modelling power, the flat (mask-based) bulk-solvent model is used by most modern crystallographic software packages to account for disordered solvent. In this model, the bulk-solvent contribution is defined by a binary mask and a scale (scattering) function. The mask is calculated on a regular grid using the atomic model coordinates and their chemical types. The grid step and two radii, solvent and shrinkage, are the three parameters that govern the mask calculation. They are highly correlated and their choice is a compromise between the computer time needed to calculate the mask and the accuracy of the mask. It is demonstrated here that this choice can be optimized using a unique value of 0.6 Å for the grid step irrespective of the data resolution, and the radii values adjusted correspondingly. The improved values were tested on a large sample of Protein Data Bank entries derived from X-ray diffraction data and are now used in the computational crystallography toolbox ( CCTBX ) and in Phenix as the default choice.

36 MATERIALS SCIENCE↗

Simulation of the Impact of Point Defects and Edge Dislocations on X-Ray Diffraction in Hexagonal (Ni,Co) 1+2 x Ti 1– x O 3 Thin Films

In this work, a computer code for simulating high-resolution X-ray diffraction (XRD) data from disordered crystals with arbitrary spatial composition and local lattice parameters is developed. Simulated patterns are compared with the experimental data collected on a single phase, highly crystalline (Ni 0.42 Co 0.58 ) 2.22 Ti 0.39 O 3 solid-solution thin film exhibiting a large number of subsidiary minima. As a case study, the edge dislocations in hexagonal (Ni,Co) 3 O 3 thin films with Burgers vector parallel and perpendicular to the a-axis and c-axis, respectively, are modeled. No peak profiles are assumed and thus issues related to profile fitting are avoided. Both macroscopic features, such as film thickness, and atomic-scale structure, such as dislocations, are simultaneously modeled. Simulations are run on a desktop machine. Commonly applied database-based phase identification routines can result in wrong phase identification, unless intensities are properly modeled. Modeling tools of diffractometer manufacturers are compared with the present approach. An example of how simulated reciprocal lattice patterns can be used to choose relevant measurement geometries in defected thin films is given.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Energy transfer and optical gain properties of P{sub 2}O{sub 5}-ZnO-LiF: (Yb{sup 3+}, Er{sup 3+}) glasses

The present paper reports on the results pertaining to the emission properties of 0.5 mol% Er{sup 3+} and together (0.5 Yb{sup 3+} /0.5 Er{sup 3+}) doped PZL (P{sub 2}O{sub 5}-ZnO-LiF) glasses prepared by a melt quenching method. From the optical absorption data, absorption and stimulated emission cross-sections have been evaluated using McCumber’s theory and further cross-sectional gain has also been computed for Yb{sup 3+}/Er{sup 3+} doped glass. On exciting the single (Er{sup 3+}) and dual rare earth ions (Yb{sup 3+}/Er{sup 3+}) doped glass sample at λ{sub exci} = 379 nm, three emission bands in the visible region {sup 2}H{sub 11/2}→{sup 4}I{sub 15/2} (526 nm), {sup 4}S{sub 3/2}→{sup 4}I{sub 15/2} (549 nm) and {sup 4}F{sub 9/2}→{sup 4}I{sub 15/2} (664 nm) are observed and while at λ{sub exci} = 980 nm (Laser Diode) excitation a broad emission at 1530 nm attributed to {sup 4}H{sub 13/2}→{sup 4}I{sub 15/2} is observed in the NIR region. The enhancement in visible and NIR emission intensities with the addition of Yb{sup 3+} to Er{sup 3+} due to an energy transfer process from Yb{sup 3+} to Er{sup 3+} has been explained in terms of an energy level diagram.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

The Electromagnetic Characteristics of the Tianlai Cylindrical Pathfinder Array

A great challenge for 21 cm intensity mapping experiments is the strong foreground radiation which is orders of magnitude brighter than the 21 cm signal. Removal of the foreground takes advantage of the fact that its frequency spectrum is smooth while the redshifted 21 cm signal spectrum is stochastic. However, a complication is the non-smoothness of the instrument response. This paper describes the electromagnetic simulation of the Tianlai cylinder array, a pathfinder for 21 cm intensity mapping experiments. Due to the vast scales involved, a direct simulation requires a large amount of computing resources. We have made the simulation practical by using a combination of methods: first simulate a single feed, then an array of feed units, finally with the feed array and a cylindrical reflector together, obtain the response for a single cylinder. We studied its radiation pattern, bandpass response and the effects of mutual coupling between feed units, and compared the results with observation. Many features seen in the measurement result are reproduced well in the simulation, especially the oscillatory features which are associated with the standing waves on the reflector. The mutual coupling between feed units is quantified with S-parameters, which decrease as the distance between the two feeds increases. Based on the simulated S-parameters, we estimate the correlated noise which has been seen in the visibility data, and the results show very good agreement with the data in both magnitude and frequency structures. These results provide useful insights on the problem of 21 cm signal extraction for real instruments.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

SU-E-T-531: Large--Scale DVH Quality Study: Correlated Aims Lead Relaxations

Purpose: Intensity modulated radiation therapy plans are designed to optimally target a tumor while sparing surrounding tissue. Desired dose distributions are iteratively approached via inverse planning. This leads to tradeoffs between clinical objectives for the planning target volume (PTV), organs at risk, and normal tissues. Dose volume histogram (DVHs) related aims are followed that are either institutional or internationally recommended. We analyze common goals and identify potential reasons that often lead to tradeoffs. Methods: 524 IMRT plans for various tumor sites were analyzed based on the main institutional DVH goal for PTV (D95) and the recommendations by ICRU-83 (D2, D50, and D98). Robust statistical tools are developed and applied to ensure that the results are immune to data uncertainties. The probability of violation was measured for each of the DVH goals based on the frequency of not meeting recommended doses. Conditional probabilities for satisfying and/or violating DVH aims were computed to test the hypothesized pair-wise relations between DVH aims. For example, for plans that satisfied D50, the probability of violating D98 was computed via P(D98 < 95% | 98% ≤ D50 ≤ 102%). The equality constraint D50 = 100% was relaxed to encompass the range [98,102]%. Results: A large majority of cases (88%) satisfied the institutional goal for PTV of D95 ≤ 95%. Similar consensus existed for D98. 51% of cases satisfied D2 ≥ 107%. However, only 18% of cases satisfied D50. The conditional probability showed correlations amongst the studied DVH goals. In fact, a negative correlation was revealed between D50 and D95 (and D98), suggesting that these competing goals cannot be satisfied concurrently. Conclusion: The majority of plans followed the institutional guidelines. The reason for their deviation from international recommendations seems to be that the latter goals are competing and cannot be satisfied concurrently in clinical practice.

60 APPLIED LIFE SCIENCES↗

FRIEDA: Flexible Robust Intelligent Elastic Data Management Framework

Scientific applications are increasingly using cloud resources for their data analysis workflows. However, managing data effectively and efficiently over these cloud resources is challenging due to the myriad storage choices with different performance, cost trade-offs, complex application choices and complexity associated with elasticity, failure rates in these environments. The different data access patterns for data-intensive scientific applications require a more flexible and robust data management solution than the ones currently in existence. FRIEDA is a Flexible Robust Intelligent Elastic Data Management framework that employs a range of data management strategies in cloud environments. FRIEDA can manage storage and data lifecycle of applications in cloud environments. There are four different stages in the data management lifecycle of FRIEDA – (i) storage planning, (ii) provisioning and preparation, (iii) data placement, and (iv) execution. FRIEDA defines a data control plane and an execution plane. The data control plane defines the data partition and distribution strategy, whereas the execution plane manages the execution of the application using a master-worker paradigm. FRIEDA also provides different data management strategies, either to partition the data in real-time, or predetermine the data partitions prior to application execution.

97 MATHEMATICS AND COMPUTING↗

Parallel beamlet dose calculation via beamlet contexts in a distributed multi‐GPU framework

Purpose Dose calculation is one of the most computationally intensive, yet essential tasks in the treatment planning process. With the recent interest in automatic beam orientation and arc trajectory optimization techniques, there is a great need for more efficient model‐based dose calculation algorithms that can accommodate hundreds to thousands of beam candidates at once. Foundational work has shown the translation of dose calculation algorithms to graphical processing units ( GPU s), lending to remarkable gains in processing efficiency. But these methods provide parallelization of dose for only a single beamlet, serializing the calculation of multiple beamlets and under‐utilizing the potential of modern GPU s. In this paper, the authors propose a framework enabling parallel computation of many beamlet doses using a novel beamlet context transformation and further embed this approach in a scalable network of multi‐ GPU computational nodes. Methods The proposed context‐based transformation separates beamlet‐local density and TERMA into distinct beamlet contexts that independently provide sufficient data for beamlet dose calculation. Beamlet contexts are arranged in a composite context array with dosimetric isolation, and the context array is subjected to a GPU collapsed‐cone convolution superposition procedure, producing the set of beamlet‐specific dose distributions in a single pass. Dose from each context is converted to a sparse representation for efficient storage and retrieval during treatment plan optimization. The context radius is a new parameter permitting flexibility between the speed and fidelity of the dose calculation process. A distributed manager‐worker architecture is constructed around the context‐based GPU dose calculation approach supporting an arbitrary number of worker nodes and resident GPU s. Phantom experiments were executed to verify the accuracy of the context‐based approach compared to Monte Carlo and a reference CPU ‐ CCCS implementation for single beamlets and broad beams composed by addition of beamlets. Dose for representative 4 π beam sets was calculated in lung and prostate cases to compare its efficiency with that of an existing beamlet‐sequential GPU ‐ CCCS implementation. Code profiling was also performed to evaluate the scalability of the framework across many networked GPU s. Results The dosimetric accuracy of the context‐based method displays <1.35% and 2.35% average error from the existing serialized CPU ‐ CCCS algorithm and Monte Carlo simulation for beamlet‐specific PDD s in water and slab phantoms, respectively. The context‐based method demonstrates substantial speedup of up to two orders of magnitude over the beamlet‐sequential GPU ‐ CCCS method in the tested configurations. The context‐based framework demonstrates near linear scaling in the number of distributed compute nodes and GPU s employed, indicating that it is flexible enough to meet the performance requirements of most users by simply increasing the hardware utilization. Conclusions The context‐based approach demonstrates a new expectation of performance for beamlet‐based dose calculation methods. This approach has been successful in accelerating the dose calculation process for very large‐scale treatment planning problems ‐ such as automatic 4 π IMRT beam orientation and VMAT arc trajectory selection, with hundreds of thousands of beamlets ‐ in clinically feasible timeframes. The flexibility of this framework makes it as a strong candidate for use in a variety of other very large‐scale treatment planning tasks and clinical workflows.

Neph, Ryan↗

A microbenchmark characterization of the Emu chick

The Emu Chick is a prototype system designed around the concept of migratory memory-side processing. Rather than transferring large amounts of data across power-hungry, high-latency interconnects, the Emu Chick moves lightweight thread contexts to near-memory cores before the beginning of each memory read. The current prototype hardware uses FPGAs to implement cache-less “Gossamer” cores for computational work and rely on a typical stationary core (PowerPC) to run basic operating system functions and migrate threads between nodes. In this multi-node characterization of the Emu Chick, we extend an earlier single-node investigation of the the memory bandwidth characteristics of the system through benchmarks like STREAM, pointer chasing, and sparse matrix-vector multiplication. Here, we compare the Emu Chick hardware to architectural simulation and an Intel Xeon-based platform. Our results demonstrate that for many basic operations the Emu Chick can use available memory bandwidth more efficiently than a more traditional, cache-based architecture although bandwidth usage suffers for computationally intensive workloads like SpMV. Moreover, the Emu Chick provides stable, predictable performance with up to 65% of the peak bandwidth utilization on a random-access pointer chasing benchmark with weak locality.

97 MATHEMATICS AND COMPUTING↗