Data Science Talk Series 

Title: From Algorithms to Accelerators: High-Performance Computing for Next-Generation Applications

Abstract: High-Performance Computing (HPC) has become a cornerstone of scientific discovery and technological advancement, enabling complex simulations, data-driven models, and large-scale computations with unprecedented speed and precision. This talk explores how modern HPC architectures and software optimizations accelerate computationally demanding applications across diverse domains, from genomic analysis to climate modeling—applications often characterized by intricate data dependencies, irregular execution patterns, and severe performance bottlenecks. We will discuss key challenges and strategies in leveraging heterogeneous computing resources, such as multi-core CPUs and GPUs, to improve efficiency and scalability. Additionally, we will highlight the role of advanced runtime systems, such as PaRSEC, in optimizing task execution and data movement for extreme-scale computations. Through case studies, we will illustrate the impact of hierarchical algorithms, efficient memory management, and scalable parallelization techniques in real-world applications. By examining state-of-the-art advancements and emerging trends, this talk aims to provide insights into how HPC continues to push the boundaries of computational feasibility, driving breakthroughs in scientific research and industrial innovation.