George Markomanolis
Principal Member of Technical Staff, AMD
Abstract
Run on AMD GPUs in minutes using ROCm: most AI frameworks offer drop-in support, while HPC applications can leverage a range of portability layers including HIP, OpenMP, Kokkos, and RAJA. Whether applications currently run on CPUs or GPUs from other vendors, the ROCm software stack provides the compilers, optimized libraries, and profiling tools needed to achieve portability and performance on AMD GPUs. In this hands-on tutorial, attendees will get access to an AMD Instinct MI300A system and learn how to port, build, and run representative HPC and AI workloads. Participants will also use ROCm profiling and performance-analysis tools to identify bottlenecks and optimization opportunities. This tutorial targets HPC and AI developers with basic or no prior GPU programming experience and provides a practical introduction to bringing existing workloads to AMD GPUs efficiently.
Biography
George Markomanolis is Principal Member of Technical Staff Software Development Engineer at AMD. He helps with the AMD training, among supporting European HPC sites. He works on understanding and porting codes for AMD GPUs and contributing to procurement benchmarking. In his current and previous role, he prepared and gave many trainings regarding HIP porting, HIP programming, benchmarking GPUs, and evaluating various programming models that can be used on AMD GPUs. His research interests are in applications porting on GPUs, benchmarking, performance evaluation/optimization of HPC applications on various technologies, and parallel I/O analysis on filesystems. Before joining AMD, he has worked in various supercomputing centers. He obtained his MSc in Computational Science from the National and Kapodistrian University of Athens, Greece in 2008 and his Ph.D. in Computer Science from the Ecole Normale Superieure de Lyon, France in 2014.
Seats are limited
Tutorials run on Monday, 26 October 2026 at KAUST. Attendance is free, but registration is required.