High-Performance GPU Programming and Sonochemistry Research Group

High-performance computing is no longer a privilege of supercomputer users. The technology has infiltrated into our everyday lives, albeit unnoticed. A personal computer with a 64-core AMD EPYC processor and two Tesla A100 graphics cards (GPUs) would have been among the top 500 supercomputers of 2010. Since GPUs account for the largest share of the performance of most high-performance workstations, one of the main activities of our research group is the application of high-performance GPU programming in the simulation of dynamical systems. The heterogeneous CPU-GPU architecture and the massively parallel nature of GPUs provoke a number of difficulties. Our research team has found effective solutions to most of the difficulties, and currently our general-purpose ordinary differential equation solver is the fastest amongst the currently available and widespread packages (www.gpuode.com). We also work on custom solutions for specific problems that fit the GPU's hardware architecture to harness most of its performance. Another major topic of our research group is a special branch of chemistry called sonochemistry. Its physical basis is the appearance of bubble clouds during the irradiation of a liquid with high intensity and frequency ultrasound. During the pulsation of the individual micron-sized bubbles, the contraction can be so high that the internal temperature can reach thousands of degrees of Kelvin, which induces chemical reactions. We are currently working on modelling and simulating the chemical processes that take place in the bubbles and bubble clouds. Our long-term goal is to design optimal rector geometries and develop optimal operating strategies for the degradation of hazardous chemical species and the synthesis of various materials (e.g., nanoparticles, ammonia). Applying high performance GPU programming is essential here.