Scalable and massively parallel Monte Carlo photon transport simulations for heterogeneous computing platforms

Area of Science:

• Computational Physics
• Scientific Computing
• Photonics

Background:

• Monte Carlo (MC) simulations are crucial for modeling photon transport.
• Existing methods often face scalability limitations on heterogeneous computing systems.
• Graphics Processing Unit (GPU)-accelerated techniques show promise but lack vendor independence.

Purpose of the Study:

• To develop a highly scalable and vendor-independent Monte Carlo simulation platform for 3D photon transport.
• To extend existing GPU-accelerated MC methods to a broader heterogeneous computing environment.
• To improve simulation performance and software portability across diverse hardware.

Main Methods:

• Development of a massively parallel Monte Carlo algorithm using the Open Computing Language (OpenCL) framework.
• Implementation of vendor-independent heterogeneous computing strategies.
• Investigation of parallel computing techniques for portable performance.
• Application of thread-level and device-level load-balancing for multi-CPU/GPU efficiency.

Main Results:

• Achieved significantly improved performance in 3D photon transport simulations.
• Demonstrated enhanced software portability across various computing hardware.
• Successfully extended MC technique to a scalable, vendor-independent heterogeneous environment.
• Validated efficient simulations using multiple central processing units (CPUs) and GPUs through advanced load balancing.

Conclusions:

• The developed platform offers a highly scalable and portable solution for 3D photon transport simulations.
• The use of OpenCL and advanced load balancing enables efficient utilization of heterogeneous computing resources.
• This research advances the field of scientific computing for complex physical phenomena.