GPU Porting for HPC: Opportunities, Tools and Libraries

Description

Abstract

 

This two-day online webinar introduces GPU porting for high-performance computing (HPC), focusing on opportunities, programming approaches and portable tools. It will help participants assess where GPU acceleration may benefit their applications and understand the first steps in adapting scientific and engineering codes.

The programme begins with an introduction to CPU and GPU architectures, their implications for performance, and the main GPU programming models. The second session introduces alpaka, an open-source C++ library for portable parallel programming across supported CPU and GPU backends. Topics include project setup, core concepts, a first kernel, and higher-level algorithms. Friday's session will start with an introduction to the multipysics framework m-AIA. 

The final session will introduce RedSim, a GPU-accelerated CFD simulation framework designed to model urban air pollution.

This course is organised by HLRS in cooperation with EXCELLERAT P2, HiDALGO2 and Plasma-PEPSC, under the umbrella of High-Performance GPU Alliance.

 

Instructors 

 

Tobias Haas (HLRS)
Tapish Narwal (HZDR / CASUS)
Tim Wegmann (RWTH Aachen University)
Mátyás Y. Constans (SZE)

 

Agenda

 

Location: Online. Hosted by HLRS. 

 

Preliminary programme. 

 

Thursday, 5. November

9:00 Welcome

9:05 - 12:00 (including breaks)

Introduction to GPU Programming - Opportunities and Challenges (HLRS)

• Differences between CPU and GPU
• GPU Architecture and Consequences
• Programming models
• Summary
• Discussion

 

13:00 - 16:00 (including breaks)

The alpaka library: task-parallel kernel execution on heterogenous hardware (PLASMA-PEPSC)

• Introduction
• Getting alpaka & project setup
• Basic concepts
• Break
• First kernel
• Higher-level algorithms & helpers
• Summar
• Discussion / Q&A

 

Friday, 6. November

9:00 - 12:00 (including breaks)

The multiphysics framework m-AIA (EXCELLERAT)

• Welcome and Introduction to the multiphysics framework m-AIA
• GPU porting in C++ with parallel STL in m-AIA
• Discussion and Coffee break
• Adaptation and dynamic load balancing on GPUs

13:00 - 16:00 (including breaks)

Efficient GPU Implementation of the Finite Volume Method on Unstructured Meshes (HIDALGO)

    • Building the Mesh
        ◦ Introduction
        ◦ Sorting as a mechanism for adjacency, RADIX Sort on the GPU
        ◦ Sub-Meshes, Halo cells
        ◦ Reordering Meshes using Space Filling Curves
        ◦ Grouping interior and boundary touching cells
    • Simulation, Runtime Communication and Kernels
        ◦ Halo communication patterns
        ◦ Maximizing work while communicationg
        ◦ GPU topolog: inter-node vs extra-node GPUs
        ◦ Load balancing
        ◦ Custom standard lobrary overview
    • Physical Method Overview, demonstration
        ◦ Solving scalar transport, kernel overviews
        ◦ Demonstration of RedSIMon a cube mesh
            ▪ SOD problem: defining the IC, BC via CUDA kernel programming
            ▪ Taylor-Green problem: defining the IC, BC, and Source term via custom CUDA kernel programming

 

Fees

This course is free of charge.

Participation is limited to people affiliated with institutions located in EU Member States or other EuroHPC JU participating countries.

For the list of participating countries, please consult the EuroHPC Joint Undertaking website.

 

Registration

 

To register, please visit the website here. 

 

Contact

 

Abdullatif Eymash

Lucienne Dettki phone +49711685 63894

 

 

Acknowledgements

This course is supported by EuroCC 3 and organised in cooperation with EXCELLERAT P2, HiDALGO2 and Plasma-PEPSC.

 

We gratefully acknowledge the funding supporting these projects:

EuroCC 3: European High-Performance Computing Joint Undertaking (EuroHPC JU) grant 101306701, supported by the European Union’s Digital Europe Programme and contributions from participating countries.

EXCELLERAT P2: EuroHPC JU grant 101092621, with funding from the European Union and Germany, Italy, Slovenia, Spain, Sweden and France.

HiDALGO2: EuroHPC JU grant 101093457, co-funded by the European Union and Poland, Germany, Spain, Hungary and France.

Plasma-PEPSC: European Union funding through grant 101093261.