Totalview in Segments

From HPC Wiki
Jump to navigation Jump to search

Introduction

Totalview is a widely used debugger in the field of HPC. This video series provides an introduction to debugging techniques and methodology, and a practical hands-on tutorial on the Totalview debugger.

This tutorial is targeted for software developers as an initial introduction to debugging with Totalview.

How to proceed through this tutorial?

The tutorial is currently made of 5 segments each detailing on basic debugging techniques. Additional segments, will be added as they are produced. Each video segments uses practical examples, provided as source files, that can be followed along. Access to the Totalview debugger is recommended to follow the tutorial along.


Who created this tutorial?

This tutorial has been developed by the Competence Center for High Performance Computing in Hessen (HKHLR).

The speaker is Dr. Christian Iwainsky from Technische Universiät Darmstadt. Christian works at the university's IT center. The slide are a collective effort of the HKHLR, and video editing was done by himself, with support by Laurin Eisenacher in editing the audio.


Videos

Each video builds on the previous videos. It is recommended to watch all videos in sequence.

Video 1: Introduction to debugging and Totalview

This video provides a brief overview of the debugging process and mindset and then introduces the debugger Totalview. Totalview Video Part I
Totalview Video - Part I

Material:
Totalview - part 1: Slides as pdf

Video 2: Steering the program control flow

This video provides basic knowledge managing control-flow in Totalview using the new user interface. Key aspects are running, pausing, stepping, advancing control-flow, run-to and step-out. Totalview Video Part II
Totalview - part 2

Material:
File:HKHLR Totalview OnlineCourseVideo2 Slides.pdf

Video 3: Inspecting variables

This video covers core-dump debugging, basic data inspection techniques and data-transformations within Totalview.

Totalview Video Part III
Totalview - part 3

Material:
Totalview - part 3: Slides as pdf

Video 4: Comfort features: action-points, and replay engine

This video covers how to attach to a running program without terminating it, break- watch- and evaluation-points, capable of interrupting execution based on configurable conditions.

Totalview Video Part IV
Totalview - part 4

Material:
Totalview - part 4: Slides as pdf

Video 5: Debugging parallel programs in Totalview

This video advances the techniques and methods from videos 1-4 into the parallel domain, covering parallel data inspection, thread and process handling, and using control groups to execute single, and groups of parallel entities.

Totalview Video Part V
Totalview - part 4

Material:
Totalview - part 5: Slides as pdf