Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “semantic data”

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 667 records · Page 37

Disentangling Rotational Dynamics and Ordering Transitions in a System of Self-Organizing Protein Nanorods via Rotationally Invariant Latent Representations

The dynamics of complex ordering systems with active rotational degrees of freedom exemplified by protein self-assembly is explored using a machine learning workflow that combines deep learning-based semantic segmentation and rotationally invariant variational autoencoder-based analysis of orientation and shape evolution. The latter allows for disentanglement of the particle orientation from other degrees of freedom and compensates for lateral shifts. The disentangled representations in the latent space encode the rich spectrum of local transitions that can now be visualized and explored via continuous variables. The time dependence of ensemble averages allows insight into the time dynamics of the system and, in particular, illustrates the presence of the potential ordering transition. Finally, analysis of the latent variables along the single-particle trajectory allows tracing these parameters on a single-particle level. The proposed approach is expected to be universally applicable for the description of the imaging data in optical, scanning probe, and electron microscopy seeking to understand the dynamics of complex systems where rotations are a significant part of the process.

representation learning↗

A combinatorial method for tracing objects using semantics of their shape.

We present a shape-first approach to finding automobiles and trucks in overhead images and include results from our analysis of an image from the Overhead Imaging Research Dataset [1]. For the OIRDS, our shape-first approach traces candidate vehicle outlines by exploiting knowledge about an overhead image of a vehicle: a vehicle's outline fits into a rectangle, this rectangle is sized to allow vehicles to use local roads, and rectangles from two different vehicles are disjoint. Our shape-first approach can efficiently process high-resolution overhead imaging over wide areas to provide tips and cues for human analysts, or for subsequent automatic processing using machine learning or other analysis based on color, tone, pattern, texture, size, and/or location (shape first). In fact, computationally-intensive complex structural, syntactic, and statistical analysis may be possible when a shape-first work flow sends a list of specific tips and cues down a processing pipeline rather than sending the whole of wide area imaging information. This data flow may fit well when bandwidth is limited between computers delivering ad hoc image exploitation and an imaging sensor. As expected, our early computational experiments find that the shape-first processing stage appears to reliably detect rectangular shapes from vehicles. More intriguing is that our computational experiments with six-inch GSD OIRDS benchmark images show that the shape-first stage can be efficient, and that candidate vehicle locations corresponding to features that do not include vehicles are unlikely to trigger tips and cues. We found that stopping with just the shape-first list of candidate vehicle locations, and then solving a weighted, maximal independent vertex set problem to resolve conflicts among candidate vehicle locations, often correctly traces the vehicles in an OIRDS scene.

99 GENERAL AND MISCELLANEOUS↗

Multidisciplinary Model Transformation Through Simplified Intermediate Representations

There has long been a challenge of making engineering tools from multiple disciplines interoperate. This problem extends to system modeling practices. This challenge has been confronted with a wide variety of techniques. These techniques include attempting to interface tools together into combined suites, attempting to find underlying commonalities in mathematics, supporting connections through semantic encoding, various graph mappings and transformations, and code wrappers. All of these approaches have strengths and weaknesses. These are measured in multiple areas: relative freedom of action of individual domain engineers in developing their own tools, speed of execution, ease of creation, traceability, fidelity of information transfer, and degree of alignment between the concepts of different domains. This paper presents an approach to this interoperation problem currently being used in the World-Wide Web. The approach is to develop easy-to-parse formats that allow flexibility to both the file author and file interpreter. Many of the formats that are currently deployed sacrifice runtime performance for the ability of third parties to easily understand what to do with the data. XML became popular earlier as a de-facto standard format for many web applications, but is now being replaced by JSON to enhance human readability and provide a simpler data model. This is the basis for work in this paper. Our approach, which provides the key to interoperation, is a simplified “shrapnel” intermediate collection of objects and relationships that is the result of a breakdown of the system model into minimal pieces. It is then reassembled on the destination side, forming a two-step transformation. Previous efforts with single-step transformations have proven too difficult to create efficiently. In contrast, the use of this approach leads to an almost automatic procedure for transformation development. The Europa project is a large engineering project that must coordinate the efforts of many different teams with different specialties. The traditional form of exchanging engineering information has been documentation. The vision of model-based systems engineering is to make this information exchange much more digital. This paper presents the application of our simplified format to connecting two different engineering tools to the system model, with a focus on a dynamic mission simulation encoded in Modelica.

Cole, Bjorn↗

Modeling Human Behavior to Anticipate Insider Attacks

The insider threat ranks among the most pressing cybersecurity challenges that threaten government and industry information infrastructures. To date, no systematic methods have been developed that provide a complete and effective approach to prevent data leakage, espionage and sabotage. Current practice is forensic in nature, relegating to the analyst the bulk of the responsibility to monitor, analyze, and correlate an overwhelming amount of data. We describe a predictive modeling framework that integrates a diverse set of data sources from the cyber domain as well as inferred psychological/motivational factors that may underlie malicious insider exploits. This comprehensive threat assessment approach provides automated support for the detection of high-risk behavioral “triggers” to help focus the analyst’s attention and inform the analysis. Designed to be domain independent, the system may be applied to many different threat and warning analysis/sensemaking problems.

99 GENERAL AND MISCELLANEOUS↗

Toward a Data Scalable Solution for Facilitating Discovery of Science Resources

Science is increasingly motivated by the need to process larger quantities of data. It is facing severe challenges in data collection, management, and processing, so much so that the computational demands of “data scaling” are competing with, and in many fields surpassing, the traditional objective of decreasing processing time. Example domains with large datasets include astronomy, biology, genomics, climate/weather, and material sciences. This paper presents a real-world use case in which we wish to answer queries pro- vided by domain scientists in order to facilitate discovery of relevant science resources. The problem is that the metadata for these science resources is very large and is growing quickly, rapidly increasing the need for a data scaling solution. We propose a system – SGEM – designed for answering graph-based queries over large datasets on cluster architectures, and we re- port performance results for queries on the current RDESC dataset of nearly 1.4 billion triples, and on the well-known BSBM SPARQL query benchmark.

aggregation↗

Evaluating MPI resource usage summary statistics

The Message Passing Interface (MPI) remains the dominant programming model for scientific applications running on today’s high-performance computing (HPC) systems. This dominance stems from MPI’s powerful semantics for inter-process communication that has enabled scientists to write applications for simulating important physical phenomena. MPI does not, however, specify how messages and synchronization should be carried out. Those details are typically dependent on low-level architecture details and the message characteristics of the application. Therefore, analyzing an application’s MPI resource usage is critical to tuning MPI’s performance on a particular platform. The result of this analysis is typically a discussion of the mean message sizes, queue search lengths and message arrival times for a workload or set of workloads. While a discussion of the arithmetic mean in MPI resource usage might be the most intuitive summary statistic, it is not always the most accurate in terms of representing the underlying data. In this paper, we analyze MPI resource usage for a number of key MPI workloads using an existing MPI trace collector and discrete-event simulator. Our analysis demonstrates that the average, while easy and efficient to calculate, is a useful metric for characterizing latency and bandwidth measurements, but may not be a good representation of application message sizes, match list search depths, or MPI inter-operation times. Additionally, we show that the median and mode are superior choices in many cases. We also observe that the arithmetic mean is not the best representation of central tendency for data that are drawn from distributions that are multi-modal or have heavy tails. Furthermore, the results and analysis of our work provide valuable guidance on how we, as a community, should discuss and analyze MPI resource usage data for scientific applications.

97 MATHEMATICS AND COMPUTING↗

Dialect Topic Modeling for Improved Consumer Medical Search

Access to health information by consumers is ham- pered by a fundamental language gap. Current attempts to close the gap leverage consumer oriented health information, which does not, however, have good coverage of slang medical terminology. In this paper, we present a Bayesian model to automatically align documents with different dialects (slang, com- mon and technical) while extracting their semantic topics. The proposed diaTM model enables effective information retrieval, even when the query contains slang words, by explicitly modeling the mixtures of dialects in documents and the joint influence of dialects and topics on word selection. Simulations us- ing consumer questions to retrieve medical information from a corpus of medical documents show that diaTM achieves a 25% improvement in information retrieval relevance by nDCG@5 over an LDA baseline.

99 GENERAL AND MISCELLANEOUS↗

Software Synthesis for High Productivity Exascale Computing

Over the three years of our project, we accomplished three key milestones: We demonstrated how ideas from generative programming and software synthesis can help support the development of bulk-synchronous distributed memory kernels. These ideas are realized in a new language called MSL, a C-like language that combines synthesis features with high level notations for array manipulation and bulk-synchronous parallelism to simplify the semantic analysis required for synthesis. We also demonstrated that these high level notations map easily to low level C code and show that the performance of this generated code matches that of handwritten Fortran. Second, we introduced the idea of solver-aided domain-specific languages (SDSLs), which are an emerging class of computer-aided programming systems. SDSLs ease the construction of programs by automating tasks such as verification, debugging, synthesis, and non-deterministic execution. SDSLs are implemented by translating the DSL program into logical constraints. Next, we developed a symbolic virtual machine called Rosette, which simplifies the construction of such SDSLs and their compilers. We have used Rosette to build SynthCL, a subset of OpenCL that supports synthesis. Third, we developed novel numeric algorithms that move as little data as possible, either between levels of a memory hierarchy or between parallel processors over a network. We achieved progress in three aspects of this problem. First we determined lower bounds on communication. Second, we compared these lower bounds to widely used versions of these algorithms, and noted that these widely used algorithms usually communicate asymptotically more than is necessary. Third, we identified or invented new algorithms for most linear algebra problems that do attain these lower bounds, and demonstrated large speed-ups in theory and practice.

97 MATHEMATICS AND COMPUTING↗

PlantSegNet: 3D point cloud instance segmentation of nearby plant organs with identical semantics

In this study, we introduce PlantSegNet, a novel neural network model for instance segmentation of nearby objects with similar geometric structures. Our work addresses the challenges of instance segmentation of plant point clouds, including the difficulty of annotating and labeling point clouds, the loss of local structural information in neural network components, and the generation of large numbers of incorrect small clusters due to poor choices of the loss function. One of the key contributions of our approach is a digital twin of sorghum, i.e., a procedural sorghum model, which was used to generate point clouds of sorghum fields. This allowed us to create a large-scale, annotated, synthetic dataset of sorghum plants that we used to train our PlantSegNet model. We demonstrated the effectiveness of our method in segmenting instances of sorghum leaves grown in outdoor field settings. To the best of our knowledge, this is the first study to address this specific instance segmentation problem for plants grown in such a setting. We compared our proposed method with other state-of-the-art methods for indoor settings, including SGPN and TreePartNet, on both synthetic and real data. Furthermore, our results show that PlantSegNet outperforms these methods regarding accuracy, robustness, and efficiency.

97 MATHEMATICS AND COMPUTING↗

GraphAide: Advanced Graph-Assisted Query and Reasoning System

Curating knowledge from multiple siloed sources that contain both structured and unstructured data is a major challenge in many real-world applications. Pattern matching and querying represent fundamental tasks in modern data analytics that leverage this curated knowledge. The development of such applications necessitates overcoming several research challenges, including data extraction, named entity recognition, data modeling, and designing query interfaces. Moreover, the explainability of these functionalities is critical for their broader adoption. The emergence of Large Language Models (LLMs) has accelerated the development lifecycle of new capabilities. Nonetheless, there is an ongoing need for domain-specific tools tailored to user activities. The creation of digital assistants has gained considerable traction in recent years, with LLMs offering a promising avenue to develop such assistants utilizing domain-specific knowledge and assumptions. In this context, we introduce an advanced query and reasoning system, GraphAide, which constructs a knowledge graph (KG) from diverse sources and allows to query and reason over the resulting KG. GraphAide harnesses both the KG and LLMs to rapidly develop domain-specific digital assistants. It integrates design patterns from retrieval augmented generation (RAG) and the semantic web to create an agentic LLM application. GraphAide underscores the potential for streamlined and efficient development of specialized digital assistants, thereby enhancing their applicability across various domains.

Purohit, Sumit [BATTELLE (PACIFIC NW LAB)] (ORCID:↗

Geospatial-Temporal Semantic Graphs for Automated Wide-Area Search

We address the problem of wide-area search of overhead imagery. Given a time sequence of overhead images, we construct a geospatial-temporal semantic graph, which expresses the complex continuous information in the overhead images in a discrete searchable form, including explicit modeling of changes seen from one image to the next. We can then express desired search goals as a template graph, and search for matches using simple and efficient graph search algorithms. This produces a set of potential matches which provide cues for where to examine the imagery in detail, applying human expertise to determine which matches are correct. We include a match quality metric that scores the matches according to how well they match the stated search goal. This enables matches to be presented in sorted order with the best matches first, similar to the results returned by a web search engine. We present an evaluation of the method applied to several examples and data sets, and show that it can be used successfully for some problems. We also remark on several limitations of the method and note additional work needed to improve its scope and robustness. Approved for public release; further dissemination unlimited.

97 MATHEMATICS AND COMPUTING↗

Regression Verification Using Impact Summaries

Regression verification techniques are used to prove equivalence of syntactically similar programs. Checking equivalence of large programs, however, can be computationally expensive. Existing regression verification techniques rely on abstraction and decomposition techniques to reduce the computational effort of checking equivalence of the entire program. These techniques are sound but not complete. In this work, we propose a novel approach to improve scalability of regression verification by classifying the program behaviors generated during symbolic execution as either impacted or unimpacted. Our technique uses a combination of static analysis and symbolic execution to generate summaries of impacted program behaviors. The impact summaries are then checked for equivalence using an o-the-shelf decision procedure. We prove that our approach is both sound and complete for sequential programs, with respect to the depth bound of symbolic execution. Our evaluation on a set of sequential C artifacts shows that reducing the size of the summaries can help reduce the cost of software equivalence checking. Various reduction, abstraction, and compositional techniques have been developed to help scale software verification techniques to industrial-sized systems. Although such techniques have greatly increased the size and complexity of systems that can be checked, analysis of large software systems remains costly. Regression analysis techniques, e.g., regression testing [16], regression model checking [22], and regression verification [19], restrict the scope of the analysis by leveraging the differences between program versions. These techniques are based on the idea that if code is checked early in development, then subsequent versions can be checked against a prior (checked) version, leveraging the results of the previous analysis to reduce analysis cost of the current version. Regression verification addresses the problem of proving equivalence of closely related program versions [19]. These techniques compare two programs with a large degree of syntactic similarity to prove that portions of one program version are equivalent to the other. Regression verification can be used for guaranteeing backward compatibility, and for showing behavioral equivalence in programs with syntactic differences, e.g., when a program is refactored to improve its performance, maintainability, or readability. Existing regression verification techniques leverage similarities between program versions by using abstraction and decomposition techniques to improve scalability of the analysis [10, 12, 19]. The abstractions and decomposition in the these techniques, e.g., summaries of unchanged code [12] or semantically equivalent methods [19], compute an over-approximation of the program behaviors. The equivalence checking results of these techniques are sound but not complete-they may characterize programs as not functionally equivalent when, in fact, they are equivalent. In this work we describe a novel approach that leverages the impact of the differences between two programs for scaling regression verification. We partition program behaviors of each version into (a) behaviors impacted by the changes and (b) behaviors not impacted (unimpacted) by the changes. Only the impacted program behaviors are used during equivalence checking. We then prove that checking equivalence of the impacted program behaviors is equivalent to checking equivalence of all program behaviors for a given depth bound. In this work we use symbolic execution to generate the program behaviors and leverage control- and data-dependence information to facilitate the partitioning of program behaviors. The impacted program behaviors are termed as impact summaries. The dependence analyses that facilitate the generation of the impact summaries, we believe, could be used in conjunction with other abstraction and decomposition based approaches, [10, 12], as a complementary reduction technique. An evaluation of our regression verification technique shows that our approach is capable of leveraging similarities between program versions to reduce the size of the queries and the time required to check for logical equivalence. The main contributions of this work are: - A regression verification technique to generate impact summaries that can be checked for functional equivalence using an off-the-shelf decision procedure. - A proof that our approach is sound and complete with respect to the depth bound of symbolic execution. - An implementation of our technique using the LLVMcompiler infrastructure, the klee Symbolic Virtual Machine [4], and a variety of Satisfiability Modulo Theory (SMT) solvers, e.g., STP [7] and Z3 [6]. - An empirical evaluation on a set of C artifacts which shows that the use of impact summaries can reduce the cost of regression verification.

Backes, John↗

Toward Usable Interactive Analytics: Coupling Cognition and Computation

Interactive analytics provide users a myriad of computational means to aid in extracting meaningful information from large and complex datasets. Much prior work focuses either on advancing the capabilities of machine-centric approaches by the data mining and machine learning communities, or human-driven methods by the visualization and CHI communities. However, these methods do not yet support a true human-machine symbiotic relationship where users and machines work together collaboratively and adapt to each other to advance an interactive analytic process. In this paper we discuss some of the inherent issues, outlining what we believe are the steps toward usable interactive analytics that will ultimately increase the effectiveness for both humans and computers to produce insights.

semantic interaction↗

Machine Learning for Predicting Team Functioning in HERA Missions

Team functioning is integral to success in future long term space exploration missions. Proactively detecting declines in team functioning can mitigate conflict and ensure mission success. This project developed a speech-based artificial intelligence (AI) system that unobtrusively predicts degradation in team functioning, including performance and cohesion, in the Human Exploration Research Analog (HERA) Campaigns 4 and 5. The AI system conducted automated analysis of the prosodic (tone of voice) and linguistic (language content) components of speech, modeling interpersonal dynamics at both the turn-taking and day-wide levels. We investigated team functioning via observing structured interactions (i.e., multi-mission space exploration vehicle-extra vehicular activity [MMSEV-EVA], team interaction battery [TIB]) and unstructured interactions before the MMSEV-EVA task. We developed machine learning models to predict team functioning (objective task accuracy, self reported team efficacy and self reported team cohesion) by analyzing OpenSmile acoustic features, linguistic descriptors extracted via the linguistic inquiry and word count (LIWC) dictionary, and semantic embeddings. In the TIB, static models using logistic regression and random forests were not able to predict task accuracy, but predicted team efficacy and cohesion during both the decision making and relational tasks to a moderate level (60-70%). Majority voting on the individual turns to predict day long team efficacy further increased accuracies (70-80%). Finally, long short-term memory (LSTM) models showed the best performance across all variables (80-91%), including task performance. In the MMSEV-EVA, static models achieved an accuracy of 60% with majority voting, which increased to 80% through the incorporation of mission day as a variable, accounting for the learning effect. A key finding across both tasks was the "team-dependent" nature of these interactions; models achieved much higher accuracy when trained on prior days of the same team's data rather than attempting to generalize across entirely different teams, with even 1-2 days of prior data per team achieving 5-15% improvement over team-independent models. In addition, the incorporation of pre-task data from the same team also improves model performance, e.g., incorporating data from the decision-making task of the TIB, which preceded the relational task, improved the prediction of team efficacy and cohesion during the latter. We compared model performance when trained on machine-generated data compared to data that had been further corrected by human annotators. Overall, models trained on human-corrected data exhibited a modest improvement in performance, particularly when acoustic features were used. We found no significant correlation between word error rate (WER) and model accuracy (r(55) = -0.08, p = 0.51), but model’s accuracy was significantly higher for medium/high quality transcription (0.74 (SD = 0.48)) compared to the low-quality group (0.64 (SD = 0.36)) (t(63)=2.82, p = 0.006). Based on these, several design recommendation emerge, that could inform Standards at NASA. Models predicting team functioning should incorporate at least one to two days of historical interaction data, include brief pre-task discussions, and explicitly model temporal learning effects, especially for longer operational tasks. Minimum quality standards for automated speech-processing pipelines are needed, given the performance gains observed with manually corrected acoustic data. Finally, systems should leverage both acoustic features and language embeddings in complementary ways, with modality choices and fusion strategies tailored to mission context, task demands, and data quality requirements.

Shrivatsa Mishra↗

Identification of threats using linguistics-based knowledge extraction.

One of the challenges increasingly facing intelligence analysts, along with professionals in many other fields, is the vast amount of data which needs to be reviewed and converted into meaningful information, and ultimately into rational, wise decisions by policy makers. The advent of the world wide web (WWW) has magnified this challenge. A key hypothesis which has guided us is that threats come from ideas (or ideology), and ideas are almost always put into writing before the threats materialize. While in the past the 'writing' might have taken the form of pamphlets or books, today's medium of choice is the WWW, precisely because it is a decentralized, flexible, and low-cost method of reaching a wide audience. However, a factor which complicates matters for the analyst is that material published on the WWW may be in any of a large number of languages. In 'Identification of Threats Using Linguistics-Based Knowledge Extraction', we have sought to use Latent Semantic Analysis (LSA) and other similar text analysis techniques to map documents from the WWW, in whatever language they were originally written, to a common language-independent vector-based representation. This then opens up a number of possibilities. First, similar documents can be found across language boundaries. Secondly, a set of documents in multiple languages can be visualized in a graphical representation. These alone offer potentially useful tools and capabilities to the intelligence analyst whose knowledge of foreign languages may be limited. Finally, we can test the over-arching hypothesis--that ideology, and more specifically ideology which represents a threat, can be detected solely from the words which express the ideology--by using the vector-based representation of documents to predict additional features (such as the ideology) within a framework based on supervised learning. In this report, we present the results of a three-year project of the same name. We believe these results clearly demonstrate the general feasibility of an approach such as that outlined above. Nevertheless, there are obstacles which must still be overcome, relating primarily to how 'ideology' should be defined. We discuss these and point to possible solutions.

99 GENERAL AND MISCELLANEOUS↗

Muscle Logic: New Knowledge Resource for Anatomy Enables Comprehensive Searches of the Literature on the Feeding Muscles of Mammals

Here we present the Mammalian Feeding Muscle Ontology (MFMO), a multi-species ontology focused on anatomical structures that participate in feeding and other oral/pharyngeal behaviors. A unique feature of the MFMO is that a simple, computable, definition of each muscle, which includes its attachments and innervation, is true across mammals. This construction mirrors the logical foundation of comparative anatomy and permits searches using language familiar to biologists. Further, it provides a template for muscles that will be useful in extending any anatomy ontology. The MFMO is developed to support the Feeding Experiments End-User Database Project (FEED, https://feedexp.org/), a publicly-available, online repository for physiological data collected from in vivo studies of feeding (e.g., mastication, biting, swallowing) in mammals. Currently the MFMO is integrated into FEED and also into two literature-specific implementations of Textpresso, a text-mining system that facilitates powerful searches of a corpus of scientific publications. We evaluate the MFMO by asking questions that test the ability of the ontology to return appropriate answers (competency questions). Lastly, we compare the results of queries of the MFMO to results from similar searches in PubMed and Google Scholar. Our tests demonstrate that the MFMO is competent to answer queries formed in the common language of comparative anatomy, but PubMed and Google Scholar are not. Overall, our results show that by incorporating anatomical ontologies into searches, an expanded and anatomically comprehensive set of results can be obtained. The broader scientific and publishing communities should consider taking up the challenge of semantically enabled search capabilities.

59 BASIC BIOLOGICAL SCIENCES↗

Benchmark Imagery Project: Selection of Real World Images for Creating Composite Images of Facilities

The "Benchmark Imagery" project is a DOE NA-22-funded effort whose three-year goal is to create a suite of imagery (both real and composite) that can be used to test geospatial algorithms used for extracting semantic content from overhead images of industrial facilities. A composite image is an image that contains both real elements such as photographic imagery and synthetic elements such as 3D models. Composite images have the advantage of being customizable and distributable, and they can contain complete ground truth. It is our goal to generate composite images which appear realistic to both human and algorithm. When generating composite images of facilities we use real overhead imagery to increase realism and decrease effort. To construct all elements with purely synthetic textures would require too much effort and still may not look as realistic as a real image. We use real time 3D rendering software "SceneWorks" for rendering composite images. In SceneWorks, a scene is constructed by placing 3D models of facility components on top of a base terrain. A base terrain is a 3D model constructed from digital elevation mapping (DEM) data and then textured with overhead orthorectified imagery. The base terrain is constructed to be geospatially accurate using a real location’s source data (DEM and Imagery.) This document outlines the selection criteria in selecting appropriate location and source data for constructing the base terrain. Imagery should have resolution equal or better than desired composite image resolution. The imagery should have no seams. Seams can occur when two or more different imagery data sets exist in a single area with neither set covering the entire area. Elevation data (DEM) resolution generally should be 1/9 arc second or better depending on the variability of elevation. The imagery must contain a flat clearing of land large enough for multiple facility types. There should be no high-rise buildings anywhere in the base terrain.

42 ENGINEERING↗

Topology-Aware Performance Optimization and Modeling of Adaptive Mesh Refinement Codes for Exascale

Here, we introduce a topology-aware performance optimization and modeling workflow for AMR simulation that includes two new modeling tools, ProgrAMR and Mota Mapper, which interface with the BoxLib AMR framework and the SSTmacro network simulator. ProgrAMR allows us to generate and model the execution of task dependency graphs from high-level specifications of AMR-based applications, which we demonstrate by analyzing two example AMR-based multigrid solvers with varying degrees of asynchrony. Mota Mapper generates multiobjective, network topology-aware box mappings, which we apply to optimize the data layout for the example multigrid solvers. While the sensitivity of these solvers to layout and execution strategy appears to be modest for balanced scenarios, the impact of better mapping algorithms can be significant when performance is highly constrained by network hop latency. Furthermore, we show that network latency in the multigrid bottom solve is the main contributing factor preventing good scaling on exascale-class machines.

97 MATHEMATICS AND COMPUTING↗