Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A GPU-accelerated simulation of rapid intensification of a tropical cyclone with observed heating

Soonpil Kang, Francis X. Giraldo, Seth Camp

TL;DR

This work delivers a GPU-accelerated, nonhydrostatic atmospheric simulation framework for tropical cyclone rapid intensification using OpenACC to port the xNUMA model. The approach combines an element-based Galerkin discretization, IMEX time integration, and tensor-based hyper-diffusion within a matrix-free, GMRES-based implicit solve, achieving energy-efficient, high-fidelity RI simulations driven by observed latent heating. Key contributions include a complete GPU/OpenACC port with explicit data management, a four-kernel RHS assembly strategy, and detailed performance analyses showing strong and weak scalability on NVIDIA A100 GPUs. Practically, the method enables faster, more energy-efficient high-resolution tropical cyclone simulations, which can inform forecasts and risk assessments. The results highlight the value of hardware-specific directive-based acceleration for scalable, high-order atmospheric modeling.

Abstract

This paper presents a limited-area atmospheric simulation of a tropical cyclone accelerated using GPUs. The OpenACC directive-based programming model is used to port the atmospheric model to the GPU. The GPU implementation of the main functions and kernels is discussed. The GPU-accelerated code produces high-fidelity simulations of a realistic tropical cyclone forced by observational latent heating. Performance tests show that the GPU-accelerated code yields energy-efficient simulations and scales well in both the strong and weak limit.

A GPU-accelerated simulation of rapid intensification of a tropical cyclone with observed heating

TL;DR

This work delivers a GPU-accelerated, nonhydrostatic atmospheric simulation framework for tropical cyclone rapid intensification using OpenACC to port the xNUMA model. The approach combines an element-based Galerkin discretization, IMEX time integration, and tensor-based hyper-diffusion within a matrix-free, GMRES-based implicit solve, achieving energy-efficient, high-fidelity RI simulations driven by observed latent heating. Key contributions include a complete GPU/OpenACC port with explicit data management, a four-kernel RHS assembly strategy, and detailed performance analyses showing strong and weak scalability on NVIDIA A100 GPUs. Practically, the method enables faster, more energy-efficient high-resolution tropical cyclone simulations, which can inform forecasts and risk assessments. The results highlight the value of hardware-specific directive-based acceleration for scalable, high-order atmospheric modeling.

Abstract

This paper presents a limited-area atmospheric simulation of a tropical cyclone accelerated using GPUs. The OpenACC directive-based programming model is used to port the atmospheric model to the GPU. The GPU implementation of the main functions and kernels is discussed. The GPU-accelerated code produces high-fidelity simulations of a realistic tropical cyclone forced by observational latent heating. Performance tests show that the GPU-accelerated code yields energy-efficient simulations and scales well in both the strong and weak limit.

Paper Structure

This paper contains 21 sections, 20 equations, 12 figures, 1 table, 3 algorithms.

Table of Contents

  1. Introduction
  2. Nonhydrostatic atmospheric model
  3. Governing equations
  4. Element-based Galerkin method
  5. Implicit-explicit (IMEX) time integration
  6. Tensor-based Hyper-Diffusion
  7. GPU implementation
  8. Base code
  9. OpenACC implementation
  10. RHS function
  11. Gradient kernel
  12. Assembly function
  13. Simulation setup
  14. Numerical setup
  15. Computing system
  16. ...and 6 more sections

Figures (12)

  • Figure 1: Layouts of domain decomposition for limited-area models. ($N_{proc,x}$ and $N_{proc,y}$ are the number of processors in the $x$ and $y$ directions.)
  • Figure 2: Simulation workflow on the CPU and GPU.
  • Figure 3: Soundings for tropical cyclone simulations: reference potential temperature $\theta_0$ and reference pressure $p_0$.
  • Figure 4: Latent heating in the region $[-60,60]\times[-60,60]\times[0,20]$ km.
  • Figure 5: Velocity magnitude at the middle $y$-$z$ plane. (The domain is vertically stretched by a factor of 10 for visual purposes.)
  • ...and 7 more figures