Stanford University logo SLAC National Accelerator Laboratory logo Los Alamos National Laboratory logo NVIDIA logo Winner of the R&D 100 Award

Legion

A Data-Centric Parallel Programming System

Github

Legion Bootcamp 2017

This page contains links to videos, slides, and exercises from the Legion Bootcamp 2017. This is the third annual meeting for people interested in learning about Legion. In a departure from the first two instances, this year’s bootcamp follows a tutorial style aimed primarily at teaching users how to program in the Legion model. The bulk of the material consists of a series of short lectures and programming exercises, progressing from “Hello World” to writing a non-trivial program capable of runnning on clusters of CPUs and GPUs. (Users coming to these materials after the fact will need to provide their own compute resources.)

The bootcamp focuses on programming in Regent, a high level, compiled language that implements the Legion programming model. Using Regent allows participants to become productive very quickly and enable the bootcamp to give participants a more in-depth understanding of the Legion approach to task-based programming. Because many projects are also interested in programming in Legion at the C++ level, there are talks in the second half of the tutorial devoted to programming directly to the Legion C++ API. The concepts in Regent and the Legion C++ API are the same, and in our experience the Regent instruction is the best way even for users primarily interested in using Legion through C++ to start.

Tutorial

Advanced Topics

Exercises

The exercises from the bootcamp are available in a Github repository. In order to run these examples on your local machine, follow the Regent installation instructions.

Previous Bootcamps