HClib: A Task-based Parallel Programming Model

December 20, 2018


Vivek Kumar, Institute of Information Technology, New Delhi, India

Vivek Sarkar, Georgia Tech., USA


Abstract: The goal of this tutorial is to help researchers and application developers to get the most their highly concurrent, heterogeneous computers using the open-sourced Habanero C/C++ library (HClib). It is a task-based parallel programming model that supports the async-finish, parallel-for, and future-promise parallel programming patterns through both C and C++ APIs. HClib explicitly exposes hardware locality of the hardware, while allowing the programmer to fall back on sane defaults. The HClib runtime is a lightweight, work-stealing, and locality-aware runtime. It is also intended to be the intra-node resource management and scheduling component within an exascale programming system, integrating with inter-node communication models such as MPI, UPC++, or OpenSHMEM.
This tutorial will provide a step-by-step introduction to using HClib (http://habanero-rice.github.io/hclib/), right from installation, setting up the environment, and to running sample parallel applications using HClib. The only pre-requisite for this tutorial is understanding of C/C++ programming languages.


How to register for this tutorial (We have limited space):

Use this link: https://docs.google.com/forms/d/17o4adaz7fXfg_NR5mcXp7O_wQfBUeIc8tb3kHqlRQt8/viewform?c=0&w=1&edit_requested=true