Real-Time Rendering of Molecular Dynamics Simulation Data: A Tutorial

Abstract

Achieving real-time molecular dynamics rendering is a challenge, especially when the rendering requires intensive computation involving a large simulation data-set. The task becomes even more challenging when the size of the data is too large to fit into random access memory (RAM) and the final imagery depends on the input and output (I/O) performance. The large data size and the complex computation processing per frame pose a number of challenges. i.e. the I/O performance bottleneck, the computational processing performance costs, and the fast rendering challenge. Handling these challenges separately consumes a significant portion of the total processing time which may result in low frame rates. We address these challenges by proposing an approach utilizing advanced memory management and bridging the Open Computing Language (OpenCL) and Open Graphics Library (OpenGL) drivers to optimize the final rendering frame rate. We illustrate the concept of the memory mapping technique and the hybrid OpenCL and OpenGL combination through a real molecular dynamics simulation example. The simulation data-set specifies the evolution of 336,260 particles over 1981 time steps occupying 8 Gigabyte of memory. The dynamics of the system including the lipid-protein interactions can be rendered at up to 40 FPS.