Molecular Dynamics (MD) Kernel Tunning for Future Hardware Nodes

Molecular dynamics (MD) simulations are pivotal in computational drug discovery, biophysics, and materials science, providing atomistic, time-resolved insights into molecular interactions. These simulations rely on computationally intensive kernels, including non-bonded force evaluations, long-range electrostatics and constraint solvers. Due to their high floating-point operation density, irregular memory access patterns, and reliance on large-scale parallelism, MD workloads place significant demands on computational and memory resources, often becoming bottlenecked by memory bandwidth and cache inefficiencies.

As hardware architectures evolve toward greater heterogeneity and specialization, it becomes essential to assess how MD kernels will perform on future systems before hardware availability. This project aims to evaluate selected MD kernels using architectural simulation platforms to analyse their performance on simulated next-generation hardware (GPU as a starting point). Through detailed analysis of architectural parameters such as memory hierarchy, SIMD width, and interconnect design and targeted tuning of system and kernel configurations, this work seeks to enable hardware-software co-design for accelerating MD simulations.

Profile: 

• Strong programming skills in C++, CUDA, ROCM, HIP
• Experience with hardware simulation platforms like gem5, AccelSim
• Exposure to benchmarking frameworks (related to GPUs) is a plus

Type of Project: Internship

Master's degree: Master of Engineering Science; Master of Engineering Technology; Master of Science

Master program: Electrotechnics/Electrical Engineering; Computer Science

Duration: 6 months

For more information or application, please contact the supervising scientist Udari De Alwis (udari.dealwis@imec.be).