TRAINING
COURSE OBJECTIVES
• Introduce HPC on distributed architecture’s fundamental principles and good practices.
• Master the use of MPI library.
• Practical application with well-known examples (research and industry).
TARGET AUDIENCE
Engineers and researchers aiming to use HPC and gain initial experience.
INSTRUCTORS
The speakers are Artelys engineers with in-depth expertise in using HPC.
PROGRAM
Presentation of High-Performance Computing
• Main issues and necessity of parallelization.
• Examples of applications.
• Hardware and software components of High-Performance Computing (processors, memory, applications allowing the implementation of parallelization on the hardware).
• Measurement of computational performance and improvement techniques.
• Introduction to parallelism performance indicators.
Architecture of HPC systems
• Shared memory architecture.
• Distributed memory architecture.
• Hybrid architectures.
High-performance architecture management systems: Scheduling and Load-balancing with SGE
Implementation of parallelization
• Concepts of message exchange. Programming interfaces.
• Parallel computing in MPI (Message Passing Interface).
– MPI-1: Number of processes, process number. Point to point communications. Collective communications. Communicators. Topologies.
– MPI-2: Dynamic process management. Parallel I/O. Memory to memory communication.
– Libraries and interfaces.
Practical work
• Getting started with MPI in Python through MPI4Py.
• Parallel simulations in MPI.
• Implementation of optimization algorithms with MPI.
PRACTICAL INFORMATION
Next session
June 2023
Training Duration
2 days
Training cost
Inter-company: €1,800 excluding taxes per person
In-house: €9,000 excluding taxes per session
This includes the materials, meals, coffee, and handouts. Computers are provided.
Artelys is a training institution, registered in France under registration number 11754066975.
Intéressé par cette formation ? Inscrivez-vous !
© ARTELYS • Tous droits réservés • Mentions légales