HiPERiSM's Course HC7 

HiPERiSM - High Performance Algorism Consulting

Course HC7: Using MPI Tools to Detect and Analyze Distributed Parallel Performance

 MPILOGOGREEN.GIF (3269 bytes) 



 
 

This training workshop can be combined with HC3 as a two-day presentation by HiPERiSM 

Prerequisites:

This course is intended for programmers who are familiar with programming for distributed memory parallel architectures with the message passing interface (MPI) in either Fortran or C. It assumes some familiarity with possible alternatives for communication methods in MPI and an appreciation of the consequences in the user’s own application.

Objectives:

This training course primarily intends to introduce the Intel Trace Analyzerô and Intel Trace Collectorô MPI programming products and teaches the methodology of detecting performance problems and correcting them using these tools. The course includes examples in Fortran and C. Detailed case studies demonstrate how these tools for MPI are applied in production models. On completion of the course participants will have the capability of understanding how performance problems in MPI code may be detected and corrected with Intel Trace Analyzerô and Intel Trace Collectorô.

Duration:

1 day organized as follows:

Day Period Chapter Topic

1

AM

 

1

2

3

Introduction

Getting started with Intel Trace Analyzerô and Intel Trace Collectorô

Using Intel Trace Analyzerô to understand performance

  PM

4

5

Control of tracing from within the source code

Performance of the Princeton Ocean Model

 

Format: 

The course is contained in a course workbook format that is intended for use in one of three ways:

  1. Class room presentation,
  2. Self-paced study,
  3. As a reference.

The workbook includes all source code and is available in Fortran, sample input, output, and make files needed to compile and execute all programs discussed in the text.

Review of Chapters:

This training workbook is arranged into four chapters described as follows.

  1. Introduction.
  2. Getting started with Intel Trace Analyzerô and Intel Trace Collectorô.
  3. Using Intel Trace Analyzerô to understand performance.
  4. Control of tracing from within the source code.
  5. Performance of the Princeton Ocean Model.

backnext page

HiPERiSM Consulting, LLC, (919) 484-9803 (Voice)

(919) 806-2813 (Facsimile)