TACCSTER 2019 Proceedings

Permanent URI for this collectionhttps://hdl.handle.net/2152/79645


Recent Submissions

Now showing 1 - 16 of 16
  • Item
    VISVE – A VIScous Vorticity Equation Model Applied to Cylinders, Hydrofoils, Propellers
    (2019) Wu, Chunlin; Kinnas, Spyros A
    A newly developed numerical tool for Computational Fluid Dynamics (CFD) computations is presented. This method is designed to be spatially compact and computationally efficient, and meanwhile capable of modeling the dynamics of vortices interacting with solid walls. The vorticity transport equation is solved in the Eulerian frame to model only the vortical regions, where the vorticity is concentrated, in the flow field. Therefore, the computational domain can be made rather small without a loss in accuracy. The small computational domain results in a significantly small number of elements in the grid and considerably less simulation time compared with the velocity based methods, such as the Navier-Stokes methods. The method has a wide range of applications, including the flow past a cylinder, sphere, hydrofoil, and propeller. In addition, both Open-MP and MPI are used to parallel the code to further facilitate its performance. Validations against the experimental results and numerical results from other methods have been conducted to verify the correctness, robustness, and computational efficiency of the current method.
  • Item
    Insight into the Sealing Capacity of Mudrocks determined using a Digital Rock Physics Workflow
    (2019) Bihani, Abhishek; Daigle, Hugh; Santos, Javier E; Landry, Christopher; Prodanović, Maša; Milliken, Kitty
    Primary objective: To better understand seal capacity in mudrocks and to determine the conditions under which a mudrock seal fails by allowing a non-wetting fluid to percolate. Hypothesis: Mudrock seals can fail below the fracture pressure if there exists a percolating pathway formed due to a continuous and sufficiently large pore-throat system. Procedure: We used SEM images of uncemented muds obtained at various depths (< 1.1 km burial) in the Kumano Basin offshore Japan for the study. Image mosaics were filtered and segmented using conventional and machine-learning techniques to identify the pore space, silt, and clay grains. We applied a 3D stochastic technique for pore space reconstruction from the SEM images and simulated capillary drainage in the resulting 3D volumes by the lattice Boltzmann method (LBM) using Stampede 2. Conclusion: Results showed that porosity and permeability decreased with depth, and capillary threshold pressure values increased. However, increasing silt content at a particular depth counteracted this behavior, due to better preservation of larger pores and throats.
  • Item
    The Path to Exascale
    (2019) Stanzione, Dan
  • Item
    Supervised Community Detection in Protein-interaction Networks
    (2019) Palukuri, Meghana; Marcotte, Edward
    Community detection problems arise in several fields with networks, from biology and medicine to social studies and cybersecurity. Networks in these fields tend to be massive - for instance hu.MAP, the human protein interaction network assembled by our lab has 17 million edges. The problem of community detection here translates to finding protein complexes, which will help advance our understanding of several cellular functions and disease mechanisms. To solve this computationally challenging big data problem, we developed Super.Complex (short for Supervised Complex), a computational pipeline employing auto-ML and subgraph sampling techniques. While most state-of-the-art algorithms employ unsupervised graph clustering methods, a supervised approach holds more promise towards finding accurate communities mimicking the real world. With data on known communities becoming increasingly available in many applications, supervised methods become more relevant. Super.Complex implements a streamlined algorithm which samples subgraphs from the weighted network and classifies them as communities or non-communities via a supervised ML model. The steps involved are (i) sampling non-community data as random walks on the graph (ii) feature extraction and selection for known communities and generated non-communities, (iii) autoML pipeline for identification and training of thebest supervised machine learning model for binary classification of subgraphs (iv) intelligent sampling of candidate subgraphs for classification via 3 search techniques – greedy, iterative simulated annealing and metropolis. The last step is in fact a solution to the NP hard problem of identifying maximally scoring subgraphs in a network. The algorithm is applied to real data of different human and yeast protein interaction networks, yielding F1 scores ranging from 0.96 to 0.99 and identifying previously unknown biological complexes. Further, Super.Complex outperforms many state-of-the-art algorithms both in terms of accuracy and performance, with scalability to huge networks through its distributed framework.
  • Item
    A High Performance Lattice Boltzmann Solver with Applications to Multiphase Flow in Porous Media
    (2019) Murakami, Margaret; Bakhshian, Sahar; Hosseini, Seyyed
    Multiphase flow is significant to many industrial processes such as the geologic storage of CO2 and oil recovery. Microscale simulation of flow in complex geological formations such as saline aquifers or oilfields is a complex and challenging task. The main goal of our study is to overcome high computational demand of multiphase flow simulations by using high performance computing. To model multiphase flow in porous media, we used a multiphase flow lattice Boltzmann (LB) method, which is recognized as an alternative to the classical computational fluid dynamics (CFD) methods. The developed LB model used an extended Color-Gradient approach with improved numerical stability, and it can be used to compute multiphase flow simulations with low capillary number and high viscosity ratios. To optimize computational efficiency, we apply the LB model to a parallel scheme written in C++ using the Message Passing Interface (MPI). Highly parallel runs of these simulations were performed using the HPC system at the Texas Advanced Computing Center at the University of Texas at Austin. We herein introduce the capability of our tool for multiphase flow simulation in porous media and present its application to CO2 sequestration in geological formations. The model has been applied to the simulation of CO2 and brine in sandstone rocks, by employing three-dimensional micro-CT images of rock samples. Injection of supercritical CO2 into the brine-saturated rock samples is simulated and complex displacement patterns under various reservoir conditions are identified.
  • Item
    Towards Real-time Fully Coherent Search for Gravitational Wave Signals from Compact Binary Coalescences
    (2019) Mohanty, Soumya; Normadin, Marc
    While a fully coherent all-sky search is known to be optimal for detecting gravitational wave signals from the inspiral and merger of a binary system of compact objects -- a compact binary coalescence (CBC) -- its high computational cost has limited current searches to less sensitive coincidence-based schemes. We have developed a code (BINARIES) that uses Particle Swarm Optimization (PSO) and an optimized numerical implementation of the mathematical formalism to speed up fully coherent all-sky search to the point where real-time analysis becomes a possibility.
  • Item
    Exploring Schrodinger’s 3-body Fermionic Wave Function in a Plane-wave Basis
    (2019) Jerke, Jonathan; Poirier, Bill
    A host of proven and new fundamental technologies allow one to approach Schrodinger’s equation in a 3-body plane-wave basis. Using the Andromeda code on Lonestar 5, my current runtime is solving lithium to chemical accuracy. I will compare and contrast pseudo-potentials and Coulomb potentials; Geminals and orbitals; and ultimately attempt the 3-electron computation.
  • Item
    Research and Development of Immersive Computational Thinking Tools using Virtual Reality, Natural Hazards Data, and Scientific Visualization to Engage K-12 Students in Scientific Computing and Engineering Education
    (2019) Garza, Edgar; Marquez, Jessica; Van Houten, Katrina
    The NHERI DesignSafe-CI Research Experience for Teachers (RET) supplement recruits two high school teachers to work alongside faculty, researchers, and staff of the Texas Advanced Computing Center (TACC) at The University of Texas at Austin. Teachers participate in graduate-level research within the fields of computing and engineering with a particular emphasis on the intersection of natural hazards data, virtual reality, and scientific visualization. The research focus in 2019 was the use of NHERI data, A Frame and WebVR framework, and TACC visualization resources to create natural hazards design features in a virtual reality environment. Professional development and training from TACC supported research deliverables, including a lesson plan aligned with Texas Essential Knowledge and Skills (TEKS) state standards, and a live demo to support TACC's education and outreach activities for K-12 and the general public. This poster will present the research process, highlighting TACC resources used, challenges and successes, and dissemination efforts.
  • Item
    Appearances of the Birthday Paradox in High Performance Computing
    (2019) Eijkhout, Victor; Myers, Margaret; McCalpin, John
    We give an elementary statistical analysis of two High Performance Computing issues, processor cache mapping and network port mapping. In both cases we find that, as in the birthday paradox, random assignment leads to more frequent coincidences than one expects a priori. Since these correspond to contention for limited resources, this phenomenon has important consequences for performance.
  • Item
    Phylogenetic and Kinematic Constraints of the Vocal-flight–respiratory Axis
    (2019) Berg, Karl
    The Motor Theory for Vocal Learning Origin posits that vocal imitation, the substrate for human speech, is a specialization of an ancestral neural pathway in the forebrain that controls locomotor activities. Birds have been the main model for understanding the biology of human speech, however, evolutionary explanations remain contentious and the Motor Theory has not been tested using phylogenetic comparative analysis, a cornerstone of evolutionary biology.
  • Item
    3D Scene Generation via Unsupervised Object Synthesis
    (2019) Beksi, William; Arshad, Mohammad Samiul
    Understanding the geometric and semantic structure of a scene (scene understanding) is a crucial problem in robotics. Researchers have employed deep learning to address scene understanding problems such as instance segmentation, semantic segmentation, and object recognition. A major impediment to applying deep learning models is the requirement for enormous quantities of labeled data: performance increases in proportion to the amount of training data available. Manually accumulating these annotated datasets is an immense undertaking and not a viable long-term option. Synthetic scene generation is an active area of research at the intersection of computer graphics, computer vision, and robotics. Recent state-of-the-art systems automatically generate configurations of objects from synthetic 3D scene models using heuristic techniques. In contrast, we introduce a framework for unsupervised synthetic scene generation from raw 3D point cloud data. Our architecture is established by autoencoders and generative adversarial networks.
  • Item
    Arctos: A Collaborative Collection Management Solution
    (2019) Zhuang, Vicky; Braker, Emily; Campbell, Mariel; Cicero, Carla; Demboski, John; Doll, Andrew; Giermakowski, Tom; Hildebrandt, Kyndall; Koo, Michelle; Linn, Angela; Mayfield-Meyer, Teresa; Spencer, Carol
    Arctos (arctosdb.org) is a cost-effective, online collaborative collection management solution employed by more than 180 collections to manage and provide access to >3.5 million biodiversity and cultural records and >775,000 media objects. It also forms the backbone of Harvard’s MCZBase. Arctos leads in providing museums with community-driven solutions to managing and improving collections data, and developing workflows for data cleaning and publication. Arctos integrates biological, earth science, cultural and emerging data types such as environmental DNA and microbiomes to provide a nexus for the full suite of object data, data derivatives and products, and their management. Arctos is used by museum professionals, researchers, educators, students, government agencies, NGOs, and the public.
  • Item
    Efficiently Mapping Linear Algebra to High-Performance Code
    (2019) Psarras, Christos; Barthels, Henrik; Bientinesi, Paolo
    Aware of the role that linear algebra plays in scientific applications, we investigate if/how matrix expressions can be efficiently evaluated with current high-level languages. On the one hand, the numerical linear algebra community has put a lot of effort in developing and optimizing a relatively small set of “universally” useful operations. These are packaged in libraries such as BLAS and LAPACK, and serve as building blocks for more complex computa- tions. On the other hand, the linear algebra expressions that arise in many domains are significantly more complex than those building blocks. We refer to the problem of expressing a linear algebra expression in terms of a set of available building blocks as the ”Linear Algebra Mapping Problem” (LAMP). In practice, users have two alternatives to solve a given LAMP: 1) either “manually”, by using C/C++ or FORTRAN in combination with explicit calls to BLAS & LAPACK 2) or “automatically” by using one of the high-level languages (or libraries) with an API that directly captures the expressions. In this presentation, we focus only on the latter. Specifically, we consider 6 languages (or libraries): Matlab, Julia, R, NumPy (Python), Eigen (C++), and Armadillo (C++), and carefully assess how effectively they translate linear algebra expressions to code, i.e., how well they solve LAMPs. We investigate a number of aspects that are critical for the efficient solution of a LAMP. These range from the most basic mapping problem “Given the expression A*B, does the language map it to a call to GEMM?”, to the optimal parenthesization, to the exploitation of properties, to the identification & elimination -if advantageous- of common sub-expressions, and more. Ultimately, the purpose of this study is to exhibit the core challenges related to the effective computation of linear algebra expressions, and to help the development of languages and libraries.
  • Item
    Trajectory Grid: A Lagrangian Advection Algorithm Implemented into CMAQ
    (2019) Pouyaei, Arman; Choi, Yunsoo; Jung, Jia; Sadeghi, Bavand
    Eulerian chemical transport models (e.g. EPA’s CMAQ) cannot clearly show the source-receptor of polluted areas. Furthermore, Lagrangian back-trajectory analysis (e.g. NOAA HYSPLIT) cannot clearly show the polluted air mass; it could be polluted or clean air mass. Trajectory Grid (TG) by Choke et al. (2005) is a Lagrangian advection algorithm. It removes the numerical diffusion caused by Eulerian advection algorithms and also transports all species at once in each “packet” so saves execution time. TG also can help in the interpretation of back-trajectories as a direct clue for aerosols variation in Eulerian models by introducing “Back Trajectory and Concentration Output.” We implemented TG into EPA’s Community Multiscale Air Quality model version 5.2.