ANACIN-X Tutorial

Non-deterministic results often arise unexpectedly in High Performance Computing (HPC) applications. These events can have a negative impact on the debugging process, and correctness of HPC and scientific simulations. ANACIN-X is a software framework specifically designed to measure the degree of non-determinism in point-to-point communication within MPI applications, utilizing graph kernel distances. This tutorial aims to assist developers and scientists in understanding and investigating the origins of non-determinism using ANACIN-X.

Prerequisites

To make most of this tutorial, it it s recommended to have a basic understanding of the following prerequisites:

  • Understanding of High Performance Computing (HPC) concepts, including parallel computing, message passing interfaces (MPI), and scientific simulations.
  • Proficiency in programming languages commonly used in HPC, such as C, C++, or Fortran.
  • Familiarity with MPI programming, including point-to-point communication, collective operations, and MPI functions.
  • Basic knowledge of graph theory concepts, such as vertices, edges, and graph connectivity
  • Familiarity with data visualization techniques, as the tutorial includes visualizing non-deterministic behavior and graph kernel distances.
  • Knowledge of Jupyter Notebook, including the ability to set up and run applications within the Jupyter environment, as the tutorial involves running ANACIN-X on Jupyter Notebook.

40 min

Tutorial Objectives

40 min

MPI-Introduction

40 min

Graphs-Introduction

40 min

ANACIN-X

40 min

ANACIN-X Steps

40 min

Running ANACIN-X

40 min

Results Interpretation

Who is the course for?

This course is designed for students, professionals and researchers in data science community. Computer science courses introduce data science students to parallel programming, but an in-depth understanding of executions on large scale HPC systems is usually not explored. Consequently, data scientists are not sufficiently trained in recognizing and addressing non-determinism when it appears in the data generation stage. Through this course, we aim to bridge the knowledge gap between the data science and HPC-enabled domain science.

About the course

This project is a joint effort between two southern academic institutions: University of Tennesse-Knoxville (UTK) and the University of North Texas (UNT).

See also

Credits

Reference