Currently a Post-doc at the National Renewable Energy Lab

Formerly a postdoc researcher at VT.

Scientific interests: 

  1. Krylov subspace methods and Krylov subspace recycling; efficiency and effectiveness of Krylov subspace solvers and preconditioners; applications of Krylov subspace methods to engineering problems, such as CFD codes, topology optimization.

  2. Scientific computing; GPU programming and code optimization.







October, 2017

Authors: Kasia Swirydowicz, Noel Chalmers, Ali Karakus, and Tim Warburton

Submitted to International Journal of HPC Applications. 

Abstract. This paper is devoted to GPU kernel optimization and performance analysis of three tensor- product operators arising in finite element methods. We provide a mathematical background to these operations and implementation details. Achieving close-to-the-peak performance for these operators requires extensive optimization because of the operators’ properties: low arithmetic intensity, tiered structure, and the need to store intermediate results inside the kernel. We give a guided overview of optimization strategies and we present a performance model that allows us to compare the efficacy of these optimizations against an empirically calibrated roofline. 




Authors: William C. Tyson, Katarzyna (Kasia) Świrydowicz, Joseph M. Derlaga, Christopher J. Roy, Eric de Sturler

Conference paper for 46th AIAA Fluid Dynamics Conference, AIAA AVIATION Forum, (AIAA 2016-3809)

Abstract: In recent years, adjoint methods have become the state of the art for both functional error estimation and adaptation. But, since most engineering applications rely upon multiple functionals to assess a physical process or system, an adjoint solution must be obtained for each functional of interest which can increase the overall computational cost significantly. In this paper, new techniques are presented which provide the same error estimates and adaptation indicators as a conventional adjoint method, but do so much more efficiently, especially when multiple functionals must be examined. For functional error estimation, the adjoint solve is replaced by the solution of an error transport equation for the local solution error and an inner product with the functional linearization. This method is shown to produce the same functional error estimate as an adjoint solve. For functional-based adaptation, the bilinear form of the functional correction is exploited to obtain the adjoint variables efficiently in an approximate sense but are still accurate enough to form useful adaptation indicators. These new methods are tested using the quasi-1D nozzle problem. Keywords: Computational fluid dynamics, Discrete adjoint method, Error estimation, Grid adaptation, Finite-volume method 



Author: Kasia Swirydowicz

Published in BIOMATH 6 (2017)

Abstract: During the 1980s and early 1990s, Martin Golubitsky and Ian Stewart  formulated and developed a theory of "coupled cell networks" (CCNs). Their research was primarily focused onquadrupeds' gaits and they applied the framework of differential equations. Golubitsky and Stewart were particularly interested in change of synchrony between $4$ legs of an animal. For example what happens when the animal speeds up from walk to gallop. 

The most important concept of their theory is a {\it cell}. The cell captures the dynamics of one unit and a dynamical system consists of many identical (governed by the same principles) cells influencing (coupling to) each other. Models based on identical cooperating units are fairly common in many areas, especially in biology, ecology and sociology. 

The goal of investigation in Coupled Cell Networks theory  is understanding the dependencies and interplay between dynamics of an individual cell, graph of connections between cells, and the nature of couplings. \vspace*{0.2em}
In this paper, I redefine Coupled Cell Networks using framework of Boolean functions. This moves the entire theory to a new setting. Some phenomena proved to be very similar as for continuous networks and some are completely different. Also, for discrete networks we ask questions differently and study different phenomena. The paper presents two examples: networks that bring 2-cell bidirectional ring as a quotient and networks that bring 3-cell bidirectional ring as a quotient.


225 Stanger St
Blacksburg, VA 24061

  • Twitter
  • LinkedIn

©2018 by the Parallel Numerical Algorithms  Research Group @VT.