After successful completion of the course, students are able to master GPU-based architectures and related technologies and to develop efficient parallel algorithms based on many-cores. Through the final assignment, they will also acquire important team working skills.

The course will start with an introduction on the modern GPU architectures, by tracing the evolution from the SIMD (Single Instruction, Multiple Data) architecture to the current architectural features and by discussing the trends for the future. We then will explore the two main Programming Models: CUDA and OpenCL. Real case studies will expose students to the potential applications of this technology. A final project will give them the possibility to make a concrete experience of the concepts taught, to solve a modest GPU programming problem that will be assigned by the teacher or proposed by the student, to present it at the end to the class. This is the list of the main topics of the course:

• GPU Architectures (NVIDIA Fermi, NVIDIA Kepler, ATI/AMD)
• CUDA Programming Model
• OpenCL Programming Model
• Case Studies (Graph exploration, Path Planning, Curvature Analysis, Signal Processing, PDE Solvers)
• Optimizing GPU performance

Registration

The enrollment can be performed using TISS. The enrollment will be close on February 20th, 2019. The max number of students for this course is 15. Please register soon !!! The first lecture will start on 3rd of March in the Seminarraum Techn. Informatik.

In the first month of the course we will present the basic knowledge that is required to do the assignment. In the second part of the course, we will present advance topics that will be useful to further improve their project. We will ask the students to setup a git hub repository where we will monitor the progress and the actual work of the single students in the group. The students will ask to make two presentations. One presentation where the students must to elaborate, before starting the project, a contingency plan where they can still safely achieve some goals and the risk is
taken into consideration. The students must write a document explaining the solution adopted for their assignment. At end of the course, the students are asked to make a public final presentation where they  their results.

ECTS-Breakdown: 6 ECTS = 150 Hours 

• 1 h - Introduction to the Course
• 1 h - Group formation
• x h - Development of project idea and brainstorming
• 20 h - Lessons on GPU Computing and Architecture
• 1 h - Workshop 1 - Project proposal presentation
• 1 h - Workshop 2 - Project results presentation
• 100 h - Project Work
•   20 h - Project discussion 

Some Resources:

• Wen-Mei W. Hwu, GPU Computing GEMS, Emerald Edition, NVIDIA
• Jason Sanders and Edward Kandrot, CUDA by Example: An Introduction to General-Purpose GPUProgramming
• Wen-mei W. Hwu, Programming Massively Parallel Processors: A Hands-on Approach
• E. Bartocci, E. Cherry, J. Glimm, R. Grosu, S.A. Smolka, and F. Fenton. Toward Real-time Simulation of Cardiac Dynamics. In CMSB 2011: Proceedings of the 9th ACM International Conference on Computational Methods in Systems Biology, Paris, France, September 21-23, pages 103-110. ACM, 2011.
• A. Murthy, E. Bartocci, F. Fenton, J. Glimm, R. Gray, S.A. Smolka, and R. Grosu. Curvature Analysis of Cardiac Excitation Wavefronts. In CMSB 2011: Proceedings of the 9th ACM International Conference on Computational Methods in Systems Biology, Paris, France, September 21-23, pages 151-160. ACM, 2011.

Students will be divided in groups of three people and a project will be assigned to each group. The project consists in solving a very high computationally intensive task, developing efficient GPU-based algorithms. At the end of the semester the students will defend their solution in a public presentation.

• VU Algorithmen und Datenstrukturen 1
• VU Algorithmen und Datenstrukturen 2
• 182.695 LU Digital Design and Computer Architecture
• 182.709 UE Operating Systems
• 351.015 VU Signals and Systems 1
• 389.055 VU Signals and Systems 2