GPU Fluid Simulation with Mimir

Overview

This project implements an incompressible fluid simulation solving the Navier-Stokes equations on the GPU using CUDA. It serves as an example and testbed for Mimir, a library for real-time visualization of GPU data. The simulation supports a variety of obstacle shapes and leverages ImGui for interactive parameter adjustments.

---

Features

• 2D incompressible Navier-Stokes solver with pressure projection
• Dynamic time step adjustment based on CFL condition
• Multiple obstacle shapes: none, square, circle, triangle, vertical and diagonal barriers
• Visualization via Mimir with buffer views updated directly from GPU memory
• Interactive GUI controls using ImGui for obstacle selection and simulation tuning

---

Requirements

• CUDA Toolkit 12.6 (or compatible)
• GCC 12.3 (or newer, supporting C++20)
• CMake 3.17 or newer
• C++20 C++20 standard enabled for host-side code
• Mimir library
• Slang shader language

---

Build Instructions

1. Create a build directory and navigate into it:

   mkdir build && cd build

2. Configure the project specifying:

• CMAKE_PREFIX_PATH: the root directory where you installed the Mimir library
• MIMIR_SHADER_DIR: the directory containing Mimir shaders

  cmake .. \
  -DCMAKE_PREFIX_PATH=/path/to/mimir_install \
  -DMIMIR_SHADER_DIR=/path/to/mimir/shaders

3. Build the project using the following command:

   make -j$(nproc)

4. Once the build completes successfully, run the simulation from the executables directory:

   ./fluid_simulation

---

Usage

• Use the GUI to:
  • Press the key V to toggle the visibility of the emitter barrier
  • Press the key F to toggle whether the obstacle follows the mouse cursor
  • Select obstacle shape (none, square, circle, triangle, vertical barrier, diagonal barrier)

Code Structure

• fluid_simulation.cu: Main simulation logic and visualization setup
• kernels.cu / kernels.cuh: CUDA kernels implementing the fluid dynamics