A Technology Training Programme: Workshop on Parallel Computing and HPC Clusters

Objective :

The workshop content forms a single concentrated course on HPC Clusters and Parallel Computing. It gives a strong foundation on programming models for Parallel Programming Platforms in the range of teraflop to petaflop computing systems. The Hands-on Session on HPC Clusters is designed to make strong foundation to understand parallel computing using different programming paradigms.

The aim is to write different example programs in numerical as well as non-numerical computations using MPI and OpenMP APIs in FORTRAN & C language focusing on different algorithms and build expertise on parallelization of serial application codes as well as analyze their performance and scalability. Besides, an overview of new trends in HPC will look into hardware accelerators and General Purpose Graphics Processing Unit (GPGPU) technologies.

The content is useful for beginners, middle level HPC Clusters and Parallel Computing users.

An overview of Parallel Computing and HPC Clusters workshop :

The workshop is focused on getting started with parallel programming using the two most widely used, highly portable and scalable Parallel Programming Models namely, MPI and OpenMP. Besides these programming models, the programme will include other aspects of Parallel Computing such as approach to write parallel codes, performance analysis of codes and new trends in parallel computing. These include the following topics :

Distributed Memory Programming with MPI (MPI-1 and MPI-2 standards)

• Derived Data types
• Group Communication
• Parallel I/O
• MPI Threads
• Dynamic Process Spawning

Shared Memory Programming with OpenMP

Case Study : Parallelization of a serial code (the serial code being developed in-house at NIO, Goa)

Parallel Algorithmic Paradigms : How to divide work among parallel processes

Profiling of codes : To identify time intensive portions in serial code that can be suitable for parallelism

Performance Metrics, Scalability and Speedup analysis of codes

Insight into Multi-core Architectures

New Trends in Parallel Computing :

GPGPU Technology : A Highly Parallel, Multithreaded and Many core Processor (Introduction to CUDA Programming - A Scalable Parallel Programming Model). The CUDA Programming Model comprises of Thread Hierarchy, Memory Hierarchy, Host and Device, Software Stack, APIs.

Hardware Accelerators - Reconfigurable Computing with FPGAs : Introduction to architecture of FPGAs, overview of VHDL and Verilog Programming Models, introduction to EDA tools.

Exposure to HPC Linux Clustering Technology : Understanding HPC Clusters and its components, Understanding Workload Managers/Schedulers, Cluster Administration.

Hands-on session : Will give an opportunity to work on various numerical and non-numerical algorithms on a Linux cluster focusing on two parallel programming models, MPI and OpenMP. The sequential and parallel programs in the hands-on session will be compiled and executed on cluster. The programming language used will be FORTRAN and C.

. . . Click to View Brochure . . .