Project Info

High Performance FDTD Computing on Apple M Chips

Mohammed Hadi
mhadi@mines.edu
Standard high performance computing platforms typically have CPU and GPU memories segregated which creates data transfer bottlenecks. FDTD, a method for electromagnetic phenomena simulations, is data heavy and computationally sparse which means it suffers heavily from this data exchange bottleneck. Apple's new M chip which is currently available on their MacBook line unifies both CPU and GPU memory banks which eliminates the need for data transfers. The resulting FDTD performance improvement is expected to be a substantial one, once FDTD codes are ported and optimized for M1 chips.

More Information:

Grand Challenge: Engineer the tools of scientific discovery
Our research group website: http://ee-arc.mines.edu/#work How to write programs for Apple GPUs: https://developer.apple.com/documentation/metal/basic_tasks_and_concepts/performing_calculations_on_a_gpu FDTD Reference: Atef Elsherbeni and Veysel Demir, The Finite Difference Time Domain Method for Electromagnetics with MATLAB Simulations, ACES Series on Computational Electromagnetics and Engineering, SciTech Publishing Inc. an Imprint of the IET, Second Edition, Edison, NJ, 2015.

Primary Contacts:

Prof. Mohammed Hadi, Electrical Engineering, mhadi@mines.edu

Student Preparation

Qualifications

Must be proficient in C or C++ or MATLAB or Fortran programming language. Knowledge of Metal Shading Language (MSL) would be a nice bonus. Physics II level understanding of electromagnetic fields. Have access to an M1 Apple MacBook Air, Pro or Mini.

TIME COMMITMENT (HRS/WK)

5

SKILLS/TECHNIQUES GAINED

Efficient programming for high performance computing platforms. Ability to write superior applications for a revolutionary computer architecture (Unified CPU and GPU memory architecture). Ability to author and present scientific posters and papers.

MENTORING PLAN

Through weekly follow-up meetings. Will teach the student the basics of FDTD. Will provide the student with ready FDTD codes (versions available in C, C++, Fortran and MATLAB, some ready for Nvidia GPU computing). Will encourage and assist the student in authoring poster and PowerPoint presentations as well journal papers, depending on research outcomes and quality of work.

PREFERRED STUDENT STATUS

Freshman
Sophomore
Junior
Senior
Share This