Roberto Giorgi
University of Siena - Dipartimento di Ingegneria dell'Informazione e Scienze Matematiche
Course Type
Type B
Calendar
September 15th, h 14/18 AULA 456 September 16th, h 9/13 AULA 456 September 17th, h 9/13 AULA 124 September 18th, h 9/13 AULA 124 September 20th, h 9/13 online (Meeting link:
https://unisi.webex.com/meet/roberto.giorgi )
Room
Program
Brief abstract
There is an ever-increasing need for more application performance and energy efficiency. However, since 2005, computing platforms have been providing substantially higher performance only by increasing the number of processors (not anymore by just increasing clock frequency).
Current processors range from multi-core encompassing, e.g., 128 cores, to GP-GPU with over 10000 cores on a single chip. These resources are aggregated via high-speed interconnects, use high-bandwidth memory, and provide a hierarchical organization of threads.
Parallel programming permits us to express an algorithm to exploit what the technology is providing us properly. But what are the principles of parallel programming that best allow (or limit) maximum performance?
This course covers several foundational topics central to parallel computing. It starts with the basic methods for properly identifying performance bottlenecks and provides the advanced practical means for programming a parallel machine using the most popular standards, such as OpenMPI, OpenMP, and CUDA. We analyze these concepts through several hands-on examples and real-world data.
Syllabus:
• Introduction and Principles of Performance Evaluation
• OpenMP, Parallel Programming for Multicore
• CUDA, Exploiting Data Parallelism
• OpenMPI, Distributing Computations across Clusters and Supercomputers
• SYCL, a new C++-based standard for Parallel Programming
All lessons will include hands-on session to familiarize with the programming paradigms and learn to write parallel programs for different modern hardware.
