HIP for CUDA Programmers, July 21, 2022

Introduction

This session is part of the OLCF’s Preparing for Frontier training series, and is open to NERSC users. 

Overview

HIP is a C++ runtime API that allows developers to write portable code to run on AMD (such as OLCF Frontier) and NVIDIA GPUs (such as NERSC Perlmutter). It is essentially a portability layer (or wrapper) that uses the underlying ROCm or CUDA platform that is installed on a system, and it is meant to have little or no performance impact over coding directly for ROCm or CUDA platforms.

In this tutorial, we will show how to get started with `hipify`-ing existing CUDA code on Summit so it can run on both NVIDIA (e.g., Summit) and ROCm platforms (e.g., Frontier). The session will consist of a main presentation with hands-on exercises throughout.

Target Audience

The target audience for this event are current Summit users who already have experience with CUDA programming and would like to start learning how to `hipify` existing CUDA applications in preparation for Frontier. The hands-on sessions will be performed on Summit (HIP is not available on Perlmutter yet) so participants without existing OLCF accounts are welcome to attend to watch the presentation, but they will not be able to participate in the hands-on sessions.

Date and Time: 10 am - 1 pm (Pacific time), Thursday, July 21, 2022

The format of this event will be online only.

Registration

Registration is required for remote participation.  Please find more information and register at the OLCF event page. 

Presentation Materials

• HIP for CUDA Programmers: slides
• HIP with Fortran: slides
• Recording (for both presentations)
• repository