Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Beatnik: A Novel Global Communication Mini-Application

Jason A. Stewart, Patrick G. Bridges

TL;DR

Beatnik introduces a lightweight open-source mini-application that stresses complex global MPI communication patterns by simulating $3$D Raleigh-Taylor instabilities via the $Z$-Model on GPU-accelerated platforms. Built with the Cabana/Kokkos performance-portability stack, it provides high-, medium-, and low-order solvers that exercise FFT-based and far-field data redistributions to reveal diverse communication patterns. The paper details the architecture, implementation, and four benchmark setups, and demonstrates scalability up to $1024$ GPUs on the Lassen system, including configurations of the HeFFTe FFT library. Together, these contributions show Beatnik’s utility for exposing MPI bandwidth/latency bottlenecks, load imbalance, and solver-library interactions, guiding performance optimizations for complex HPC workloads.

Abstract

Beatnik is a novel open source mini-application that exercises the complex communication patterns often found in production codes but rarely found in benchmarks or mini-applications. It simulates 3D Raleigh-Taylor instabilities based on Pandya and Shkoller's Z-Model formulation using the Cabana performance portability framework. This paper presents both the high-level design and important implementation details about Beatnik along with four benchmark setups for evaluating different aspects of HPC communication system performance. Evaluation results demonstrate both Beatnik's scalability on modern accelerator-based systems using weak and strong scaling tests up to 1024 GPUs, along with Beatnik's ability to expose communication challenges in modern systems and solver libraries.

Beatnik: A Novel Global Communication Mini-Application

TL;DR

Beatnik introduces a lightweight open-source mini-application that stresses complex global MPI communication patterns by simulating D Raleigh-Taylor instabilities via the -Model on GPU-accelerated platforms. Built with the Cabana/Kokkos performance-portability stack, it provides high-, medium-, and low-order solvers that exercise FFT-based and far-field data redistributions to reveal diverse communication patterns. The paper details the architecture, implementation, and four benchmark setups, and demonstrates scalability up to GPUs on the Lassen system, including configurations of the HeFFTe FFT library. Together, these contributions show Beatnik’s utility for exposing MPI bandwidth/latency bottlenecks, load imbalance, and solver-library interactions, guiding performance optimizations for complex HPC workloads.

Abstract

Beatnik is a novel open source mini-application that exercises the complex communication patterns often found in production codes but rarely found in benchmarks or mini-applications. It simulates 3D Raleigh-Taylor instabilities based on Pandya and Shkoller's Z-Model formulation using the Cabana performance portability framework. This paper presents both the high-level design and important implementation details about Beatnik along with four benchmark setups for evaluating different aspects of HPC communication system performance. Evaluation results demonstrate both Beatnik's scalability on modern accelerator-based systems using weak and strong scaling tests up to 1024 GPUs, along with Beatnik's ability to expose communication challenges in modern systems and solver libraries.
Paper Structure (18 sections, 9 figures, 1 table)

This paper contains 18 sections, 9 figures, 1 table.

Table of Contents

  1. Introduction
  2. Benchmark Goals and Architecture
  3. Implementation
  4. Core Modules
  5. Birchoff-Rott Solvers
  6. Exact Integral Calculation
  7. Cutoff-based Approximate Integral Calculation
  8. Mini-Application Driver Problem and Input Decks
  9. Benchmark Evaluation
  10. Experimental Setup
  11. Low-Order Solver Weak and Strong Scaling
  12. High-Order Cutoff Solver Weak Scaling
  13. Strong Scaling of the Cutoff Solver
  14. Evaluation of HeFFTe Configurations
  15. Future Work
  16. ...and 3 more sections

Figures (9)

  • Figure 1: 4-GPU multi-mode test case output using low-order solver at timestep 20 with mesh points colored by vorticity magnitude
  • Figure 2: 36-GPU single-mode test case output using the cutoff solver at timestep 340 with mesh points colored by vorticity magnitude
  • Figure 3: Low-Order Weak Scaling of Beatnik on Lassen
  • Figure 4: Low-Order Strong Scaling of Beatnik on Lassen
  • Figure 5: Total Runtime When Weak-Scaling the Cutoff Solver
  • ...and 4 more figures