Milinda Fernando
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Milinda Fernando has a total of 50 co-authors. Here are the top co-authors by number of shared publications:
- Hari Sundar
Department of Computer Science, Tufts University, Medford, MA, USA
8 shared publications - David Neilsen
Department of Physics and Astronomy Brigham Young University 84602 Provo UT
4 shared publications Hari Sundar
University of Utah
4 shared publications- Kumar Saurabh
Department of Mechanical Engineering Iowa State University Ames IA
3 shared publications - Baskar Ganapathysubramanian
Department of Mechanical Engineering, Iowa State University, Ames, USA
2 shared publications David Neilsen
Brigham Young University
2 shared publications
Latest Publications
- 13 October 2026
- Published by Elsevier in Computer Physics Communications
- Vol. 330
A batched electron-Boltzmann solver for low-temperature plasmas
- 13 October 2026
- Published by Elsevier in Computer Physics Communications
- Vol. 330
A batched electron-Boltzmann solver for low-temperature plasmas
- Open Access
- 15 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 60-71
Real-Time Bayesian Inference at Extreme Scale: A Digital Twin for Tsunami Early Warning Applied to the Cascadia Subduction Zone
- 15 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 60-71
Open AccessReal-Time Bayesian Inference at Extreme Scale: A Digital Twin for Tsunami Early Warning Applied to the Cascadia Subduction Zone
- 9 October 2025
- Published by Society for Industrial & Applied Mathematics (SIAM) in SIAM Journal on Scientific Computing
- Vol. 47 (5) , B1201-B1226
Fast and Scalable FFT-Based GPU-Accelerated Algorithms for Block-Triangular Toeplitz Matrices with Application to Linear Inverse Problems Governed by Autonomous Dynamical Systems
We present an efficient and scalable algorithm for performing matrix-vector multiplications (matvecs) for block Toeplitz matrices. Such matrices, which are shift-invariant with respect to their blocks, arise in the context of solving inverse problems governed by autonomous systems, and time-invariant systems in particular. In this article, we consider inverse problems that infer unknown parameters from observational data of a linear time-invariant dynamical system given in the form of partial differential equations (PDEs). Matrix-free Newton-conjugate-gradient methods are often the gold standard for solving these inverse problems, but they require numerous actions of the Hessian on a vector. Matrix-free adjoint-based Hessian matvecs require solution of a pair of linearized forward/adjoint PDE solves per Hessian action, which may be prohibitive for large-scale inverse problems. Time invariance of the forward PDE problem leads to a block Toeplitz structure of the discretized parameter-to-observable (p2o) map defining the mapping from inputs (parameters) to outputs (observables) of the PDEs. This block Toeplitz structure enables us to exploit two key properties: (1) compact storage of the p2o map and its adjoint, and (2) efficient fast Fourier transform–based Hessian matvecs. The proposed algorithm is mapped onto large multi-GPU clusters and achieves more than 80% of peak bandwidth on NVIDIA A100 GPUs. Excellent weak scaling is shown for up to 48 A100 GPUs. For the targeted problems, the implementation executes Hessian matvecs within fractions of a second, which is orders of magnitude faster than can be achieved by conventional matrix-free Hessian matvecs via forward/adjoint PDE solves. Reproducibility of computational results. This paper has been awarded the “SIAM Reproducibility Badge: Code and data available” as a recognition that the authors have followed reproducibility principles valued by SISC and the scientific computing community. Code and data that allow readers to reproduce the results in this paper are available at https://github.com/s769/FFTMatvec and in the supplementary materials ( FFTMatvec-main.zip [1.05MB]). [Formula: see text]
- 9 October 2025
- Published by Society for Industrial & Applied Mathematics (SIAM) in SIAM Journal on Scientific Computing
- Vol. 47 (5) , B1201-B1226
Fast and Scalable FFT-Based GPU-Accelerated Algorithms for Block-Triangular Toeplitz Matrices with Application to Linear Inverse Problems Governed by Autonomous Dynamical Systems
We present an efficient and scalable algorithm for performing matrix-vector multiplications (matvecs) for block Toeplitz matrices. Such matrices, which are shift-invariant with respect to their blocks, arise in the context of solving inverse problems governed by autonomous systems, and time-invariant systems in particular. In this article, we consider inverse problems that infer unknown parameters from observational data of a linear time-invariant dynamical system given in the form of partial differential equations (PDEs). Matrix-free Newton-conjugate-gradient methods are often the gold standard for solving these inverse problems, but they require numerous actions of the Hessian on a vector. Matrix-free adjoint-based Hessian matvecs require solution of a pair of linearized forward/adjoint PDE solves per Hessian action, which may be prohibitive for large-scale inverse problems. Time invariance of the forward PDE problem leads to a block Toeplitz structure of the discretized parameter-to-observable (p2o) map defining the mapping from inputs (parameters) to outputs (observables) of the PDEs. This block Toeplitz structure enables us to exploit two key properties: (1) compact storage of the p2o map and its adjoint, and (2) efficient fast Fourier transform–based Hessian matvecs. The proposed algorithm is mapped onto large multi-GPU clusters and achieves more than 80% of peak bandwidth on NVIDIA A100 GPUs. Excellent weak scaling is shown for up to 48 A100 GPUs. For the targeted problems, the implementation executes Hessian matvecs within fractions of a second, which is orders of magnitude faster than can be achieved by conventional matrix-free Hessian matvecs via forward/adjoint PDE solves. Reproducibility of computational results. This paper has been awarded the “SIAM Reproducibility Badge: Code and data available” as a recognition that the authors have followed reproducibility principles valued by SISC and the scientific computing community. Code and data that allow readers to reproduce the results in this paper are available at https://github.com/s769/FFTMatvec and in the supplementary materials ( FFTMatvec-main.zip [1.05MB]). [Formula: see text]
- 3 June 2025
- Published by American Physical Society (APS) in Physical Review D
- Vol. 111 (12)
Nyquist-resolving gravitational waves via orbital frequency-based refinement
- 3 June 2025
- Published by American Physical Society (APS) in Physical Review D
- Vol. 111 (12)
Nyquist-resolving gravitational waves via orbital frequency-based refinement
- Open Access
Boltzsim: A fast solver for the 1D-space electron Boltzmann equation with applications to radio-frequency glow discharge plasmas
We present an algorithm for solving the one-dimensional space collisional Boltzmann transport equation (BTE) for electrons in low-temperature plasmas (LTPs). Modeling LTPs is useful in many applications, including advanced manufacturing, material processing, and hypersonic flows, to name a few. The proposed BTE solver is based on an Eulerian formulation. It uses Chebyshev collocation method in physical space and a combination of Galerkin and discrete ordinates in velocity space. We present self-convergence results and cross-code verification studies compared to an in-house particle-in-cell (PIC) direct simulation Monte Carlo (DSMC) code. Boltzsim is our open source implementation of the solver. Furthermore, we use Boltzsim to simulate radio-frequency glow discharge plasmas (RF-GDPs) and compare with an existing methodology that approximates the electron BTE. We compare these two approaches and quantify their differences as a function of the discharge pressure. The two approaches show an 80x, 3x, 1.6x, and 0.98x difference between cycle-averaged time periodic electron number density profiles at 0.1 Torr, 0.5 Torr, 1 Torr, and 2 Torr discharge pressures, respectively. As expected, these differences are significant at low pressures, for example less than 1 Torr.
Open AccessBoltzsim: A fast solver for the 1D-space electron Boltzmann equation with applications to radio-frequency glow discharge plasmas
We present an algorithm for solving the one-dimensional space collisional Boltzmann transport equation (BTE) for electrons in low-temperature plasmas (LTPs). Modeling LTPs is useful in many applications, including advanced manufacturing, material processing, and hypersonic flows, to name a few. The proposed BTE solver is based on an Eulerian formulation. It uses Chebyshev collocation method in physical space and a combination of Galerkin and discrete ordinates in velocity space. We present self-convergence results and cross-code verification studies compared to an in-house particle-in-cell (PIC) direct simulation Monte Carlo (DSMC) code. Boltzsim is our open source implementation of the solver. Furthermore, we use Boltzsim to simulate radio-frequency glow discharge plasmas (RF-GDPs) and compare with an existing methodology that approximates the electron BTE. We compare these two approaches and quantify their differences as a function of the discharge pressure. The two approaches show an 80x, 3x, 1.6x, and 0.98x difference between cycle-averaged time periodic electron number density profiles at 0.1 Torr, 0.5 Torr, 1 Torr, and 2 Torr discharge pressures, respectively. As expected, these differences are significant at low pressures, for example less than 1 Torr.
- 17 March 2023
- Published by American Physical Society (APS) in Physical Review D
- Vol. 107 (6) , 064035
Massively parallel simulations of binary black holes with adaptive wavelet multiresolution
We present results from the new dendro-gr code. These include simulations of binary black hole mergers for mass ratios up to . dendro-gr uses wavelet adaptive multiresolution to generate an unstructured grid adapted to the spacetime geometry together with an octree-based data structure. We demonstrate good scaling, improved convergence properties, and efficient use of computational resources. We validate the code with comparisons to lazev.- 17 March 2023
- Published by American Physical Society (APS) in Physical Review D
- Vol. 107 (6) , 064035
Massively parallel simulations of binary black holes with adaptive wavelet multiresolution
We present results from the new dendro-gr code. These include simulations of binary black hole mergers for mass ratios up to . dendro-gr uses wavelet adaptive multiresolution to generate an unstructured grid adapted to the spacetime geometry together with an octree-based data structure. We demonstrate good scaling, improved convergence properties, and efficient use of computational resources. We validate the code with comparisons to lazev.- Open Access
Massively parallel simulations of binary black holes with Dendro-GR
We present results from the new Dendro-GR code. These include simulations of binary black hole mergers for mass ratios up to q=16. Dendro-GR uses Wavelet Adaptive Multi-Resolution (WAMR) to generate an unstructured grid adapted to the spacetime geometry together with an octree based data structure. We demonstrate good scaling, improved convergence properties and efficient use of computational resources. We validate the code with comparisons to LazEv.
Open AccessMassively parallel simulations of binary black holes with Dendro-GR
We present results from the new Dendro-GR code. These include simulations of binary black hole mergers for mass ratios up to q=16. Dendro-GR uses Wavelet Adaptive Multi-Resolution (WAMR) to generate an unstructured grid adapted to the spacetime geometry together with an octree based data structure. We demonstrate good scaling, improved convergence properties and efficient use of computational resources. We validate the code with comparisons to LazEv.
- Open Access
A fully-coupled framework for solving Cahn-Hilliard Navier-Stokes equations: Second-order, energy-stable numerical methods on adaptive octree based meshes
We present a fully-coupled, implicit-in-time framework for solving a thermodynamically-consistent Cahn-Hilliard Navier-Stokes system that models two-phase flows. In this work, we extend the block iterative method presented in Khanwale et al. [\textit{Simulating two-phase flows with thermodynamically consistent energy stable Cahn-Hilliard Navier-Stokes equations on parallel adaptive octree based meshes}, J. Comput. Phys. (2020)], to a fully-coupled, provably second-order accurate scheme in time, while maintaining energy-stability. The new method requires fewer matrix assemblies in each Newton iteration resulting in faster solution time. The method is based on a fully-implicit Crank-Nicolson scheme in time and a pressure stabilization for an equal order Galerkin formulation. That is, we use a conforming continuous Galerkin (cG) finite element method in space equipped with a residual-based variational multiscale (RBVMS) procedure to stabilize the pressure. We deploy this approach on a massively parallel numerical implementation using parallel octree-based adaptive meshes. We present comprehensive numerical experiments showing detailed comparisons with results from the literature for canonical cases, including the single bubble rise, Rayleigh-Taylor instability, and lid-driven cavity flow problems. We analyze in detail the scaling of our numerical implementation.
Open AccessA fully-coupled framework for solving Cahn-Hilliard Navier-Stokes equations: Second-order, energy-stable numerical methods on adaptive octree based meshes
We present a fully-coupled, implicit-in-time framework for solving a thermodynamically-consistent Cahn-Hilliard Navier-Stokes system that models two-phase flows. In this work, we extend the block iterative method presented in Khanwale et al. [\textit{Simulating two-phase flows with thermodynamically consistent energy stable Cahn-Hilliard Navier-Stokes equations on parallel adaptive octree based meshes}, J. Comput. Phys. (2020)], to a fully-coupled, provably second-order accurate scheme in time, while maintaining energy-stability. The new method requires fewer matrix assemblies in each Newton iteration resulting in faster solution time. The method is based on a fully-implicit Crank-Nicolson scheme in time and a pressure stabilization for an equal order Galerkin formulation. That is, we use a conforming continuous Galerkin (cG) finite element method in space equipped with a residual-based variational multiscale (RBVMS) procedure to stabilize the pressure. We deploy this approach on a massively parallel numerical implementation using parallel octree-based adaptive meshes. We present comprehensive numerical experiments showing detailed comparisons with results from the literature for canonical cases, including the single bubble rise, Rayleigh-Taylor instability, and lid-driven cavity flow problems. We analyze in detail the scaling of our numerical implementation.
- 13 April 2022
- Published by Society for Industrial & Applied Mathematics (SIAM) in SIAM Journal on Scientific Computing
- Vol. 44 (2) , C156-C183
Scalable Local Timestepping on Octree Grids
Numerical solutions of hyperbolic partial differential equations (PDEs) are ubiquitous in science and engineering. Method of lines is a popular approach to discretize PDEs defined in spacetime, where space and time are discretized independently. When using explicit timesteppers on adaptive grids, the use of a global timestep size dictated by the finest grid-spacing leads to inefficiencies in the coarser regions. Even though adaptive space discretizations are widely used in computational sciences, temporal adaptivity is less common due to its sophisticated nature. This paper presents highly scalable algorithms to enable local timestepping (LTS) for explicit timestepping schemes on octrees. We demonstrate the accuracy of our methods and the scalability of our framework across 16K cores on TACC's Frontera supercomputer https://frontera-portal.tacc.utexas.edu/. We also present a speedup estimation model for LTS that predicts the distributed speedup compared to global timestepping with an average of 0.1 relative error.
- 13 April 2022
- Published by Society for Industrial & Applied Mathematics (SIAM) in SIAM Journal on Scientific Computing
- Vol. 44 (2) , C156-C183
Scalable Local Timestepping on Octree Grids
Numerical solutions of hyperbolic partial differential equations (PDEs) are ubiquitous in science and engineering. Method of lines is a popular approach to discretize PDEs defined in spacetime, where space and time are discretized independently. When using explicit timesteppers on adaptive grids, the use of a global timestep size dictated by the finest grid-spacing leads to inefficiencies in the coarser regions. Even though adaptive space discretizations are widely used in computational sciences, temporal adaptivity is less common due to its sophisticated nature. This paper presents highly scalable algorithms to enable local timestepping (LTS) for explicit timestepping schemes on octrees. We demonstrate the accuracy of our methods and the scalability of our framework across 16K cores on TACC's Frontera supercomputer https://frontera-portal.tacc.utexas.edu/. We also present a speedup estimation model for LTS that predicts the distributed speedup compared to global timestepping with an average of 0.1 relative error.