Home>> | Back to Courseware Access page >> | Back to Other Courseware page >> |

Link to the advanced 2. High Performance Computing MPI Programming section >> 

1. Introduction to High-Performance Computing

Operating System: Linux*

Language: English

Audience: Cluster Users and Developers

Courseware Description:

These modules form an intensive two-day overview, with hands-on lab exercises, basics of high-performance cluster computing, use of cluster resources and application preparation. This courseware also includes an overview of performance analysis and tuning of multi-threaded and parallel applications, as well as application debugging and porting tips.

Course Objectives:

After completing this course, you should be able to:

• Be able to understand and run cluster applications;
• Prepare your own cluster ready application;
• Basic cluster application optimizations;
• Diagnose application stability problems;
• Understand challenges and have basic recipes for application porting.

HPC Course Module Overview:

H.1.a Introduction to High Performance Cluster computing module (1 hour)

H.1.b Preparing applications to run on a cluster module (3 hours)

H.1.c Tuning HPC applications (3 hours)

H.1.e Tips for porting to 64-bit (Intel Processors) (1.5 hours)

H.1.a Introduction to High Performance Cluster computing module (1 hour)

Objectives:

• Introduce basic concepts of cluster computing, evolution of parallel processing architectures
• Describe hardware and software components of a cluster
• Applications and usage models: Fine-Grained, Coarse-Grained, Embarrassingly Parallel Applications

Objectives:

• Use key compiler optimization switches
• Link with standard libraries: BLAS, LAPACK, FFTW, etc
• Link with MPI libraries
• Creating PBSpro job submission script
  • Running large number of serial (non-MPI) jobs
  • Running MPI jobs
  • Running heterogeneous applications supporting MPI and OpenMP (ppn)

Agenda:

• Compiler Switches
• General Fortran & C++ switches
• Optimized Mathematic Libraries
• Introduction of OpenMP
• MPI libraries
• Job scheduling system
• Running the jobs HOWTOs
• Hands-on

Objectives:

• Understand and identify bottlenecks
• Apply tuning methodologies

Agenda:

• Application efficiency: scaling and tuning concepts
• Monitoring application performance (? – Darwin)
• Overview of parallel algorithms
• Multithreaded Programming Methodology
• MPI tuning

Objectives:

• Understand 64-bit porting concepts and issues

Agenda:

• Port applications to 64-bit
• Fighting copmpiler compatibility / compliance issues
• MPI compliance issues

• Preparing applications to run on a cluster module (3 hours)
• Tuning HPC applications module (3 hours)
• Debugging HPC applications module (2 hours)
• Porting tips to 64-bit Intel processors module (1.5 hours)