CUDA Spotlight: GPU-Accelerated Electromagnetic and Micromagnetic Simulators

Return to CUDA Spotlight Archive

This week’s spotlight is on Vitaliy Lomakin, an Associate Professor at the University of California, San Diego (UCSD). Prof. Lomakin works in the field of Computational Electromagnetics and Micromagnetics. This technology is fundamental in the design of components in products like solar cells, antennas and hard drives. In addition, it is utilized in R&D related to human health, from MRI systems to treatments for cancer.

NVIDIA: Vitaliy, please tell us about what you and your team are working on.
Vitaliy: My research group develops high-performance electromagnetic simulators, which solve Maxwell’s equations, and micromagnetic simulators, which solve the Landau-Lifshitz-Gilbert equation. These simulators have a significant predictive power - allowing the analysis, design, and validation of systems before and during fabrication and testing.

Electromagnetic simulators are used in the areas of optical systems (solar cells, lasers, waveguides, fibers, sensors, optical imaging systems, etc.) and microwave components (antennas, radars, communication systems, microwave imaging systems, etc.).

Micromagnetic simulators are critical in the research and development of magnetic resonance imaging systems, magnetic hyperthermia systems, magnetic recording systems, spintronics, and magnetic memories, to name a few. As an example, our simulators have enabled - for the first time ever - the modeling of the highly complex problem of a complete magnetic recording head used in hard drives, which needs to be discretized over hundreds of millions of elements.

Magnetization state in a recording head, which was discretized into 126 million tetrahedrons.

NVIDIA: When did you start using GPUs? How does GPU computing impact your work?
Vitaliy: I first read about GPUs around three years ago and we decided to give it a try. A Ph.D. student in my group, Shaojing Li, took on the project and did superb work. Now the majority of our codes use GPUs.

Our simulators are implemented on GPU-based systems using the CUDA parallel programming model. As a result, they are very efficient in terms of speedups (70-400X, as compared to CPUs) and absolute performance. For example, we developed a micromagnetic solver, referred to as FastMag. On a simple desktop, FastMag can rapidly solve complex problems discretized over hundreds of millions tetrahedral elements. This is enabled by GPU computing.

The advantages of working with CUDA include the ability to handle very large problem sizes; very fast performance; easily obtainable computers; no need for specialized facilities for computers; and low power consumption. It offers a truly high-performance cluster at a cost of a desktop.

NVIDIA: What sparked your interest in electromagnetics?
Vitaliy: My father was an electrical engineer and two of my uncles are professors. From an early age I was exposed to physics and math. In my undergraduate school, I found the program offered by the Department of Theoretical Radio Physics to be the most challenging and interesting. I chose to go with this department, started doing research on electromagnetics, and have been doing it ever since.

Simulations of spin waves excited in a thin magnetic film by a spin transfer torque nano-oscillator.

NVIDIA: As computing becomes more powerful, what will we be able to do in the future?
Vitaliy: I believe GPU computing will play an essential role in our ability to model and - most importantly - to design and optimize complete magnetic, microwave and optical systems (not only system components). An exciting example is heat-assisted magnetic recording (HAMR) systems, which are envisioned as the next hard drive technology. In HAMR, localized optical fields are used to locally heat magnetic media, assisting magnetic recording. Such systems are complex to analyze. High-performance computational tools will allow complete systems analysis with discretization on the order of a billion elements, both for optics and magnetics designs.

NVIDIA: Is there anything else you would like to tell us?
Vitaliy: I am an Associate Editor for the International Journal of Numerical Modelling: Network, Devices and Fields. Realizing the importance of GPUs, and believing in the future of GPU computing, I am organizing a Special Issue on GPU computing for the journal. The issue title is “High-Performance Computing on Graphics Processing Units for Field and Device Modelling.” We recently opened up the Call for Papers. You can submit a paper here.

Additionally, my group is involved in several computational projects, including the Center for Magnetic Recording Research (CMRR) at UCSD and the NSF ERC Center for Integrated Access Networks (CIAN).

With CMRR, a world-renowned center in the field of magnetic recording, we work with hard drive companies to advance ultra-high density magnetic recording. With CIAN, a research leader in future optical systems, we develop computational electromagnetics tools in the framework of the Optical SPICE project to enable microsystems on a chip integrating photonic and electronic functionalities for future applications.

Prof. Lomakin’s Bio:
Prof. Lomakin received his M.S. from Kharkiv National University (Ukraine) and Ph.D. from Tel Aviv University (Israel), both in Electrical Engineering. Formerly, he was a Visiting Asst. Professor at the University of Illinois at Urbana-Champaign. At UCSD, he teaches undergraduate and graduate courses and conducts research.

You can read more about the work of Prof. Lomakin and his colleagues in these articles:

1. R. Chang, S. Li, M. Lubarda, B. Livshitz, V. Lomakin, “FastMag: Fast micromagnetic simulator for complex magnetic structures,” J. Appl. Phys., in press (invited).
2. S. Li, B. Livshitz, V. Lomakin, “Fast evaluation of Helmholtz potential on graphics processing units (GPUs),” Journal of Computational Physics, vol. 229, pp. 8463-8483, 2010.
3. S. Li, B. Livshitz, V. Lomakin, “Graphics Processing Unit accelerated O(N) micromagnetic solver,” IEEE Transactions on Magnetics, vol. 46, no. 6, pp. 2373 - 2375, 2010.