Fluid Simulation
December 4, 2019 ยท View on GitHub
LBM fluid simulation based on wgpu-rs.
src/lbm implementation is standard LBM.
src/optimized_mem_lbm implementation is based on the following paper:
A New Approach to Reduce Memory Consumption in Lattice Boltzmann Method on GPU
Performance per frame
render size: full screen on iPhone, 1600px x 1200px on macOS.
Standard LBM Poiseuille flow with 100 * 75 lattice
| Device | OS Version | FPS | CPU cost | GPU frame time |
|---|---|---|---|---|
| MBP 2018 (i7, AMD Radeon Pro 560X) | macOS 10.15.1 | 60fps | 17% | 3.8ms |
| MBP 2018 (i7, Intel(R) UHD Graphics 630) | macOS 10.15.1 | 60fps | 21% | 2.0ms |
| iPhone 6 plus | iOS 12.4.3 | 60fps | 80% | 13.5ms |
| iPhone Xs Max | iOS 13.2.3 | 60fps | 54% | 6.3ms |
| iPad Pro (11-inch) | iOS 13.1.3 | 12fps | 83% | 5.4ms |
Memory consumption optimized Poiseuille flow with 200 * 150 lattice and particles separately
| Device | OS Version | FPS | Memory | CPU cost | GPU frame time |
|---|---|---|---|---|---|
| iPhone 6 plus | iOS 12.4.3 | 60fps | 88.8 MB | 88% | 16.2ms |
| iPhone Xs Max | iOS 13.2.3 | 60fps | 148.7 MB | 53% | 8.2ms |
| iPad Pro (11-inch) | iOS 13.1.3 | 120fps | 165.5 MB | 87% | 6.2ms |
Run
# on macOS
# run poiseuille flow
cargo run --example fluid poiseuille
# run lid-driven cavity
cargo run --example fluid lid-driven-cavity
# run pigments diffuse
cargo run --example fluid pigments-diffuse
# build for iOS
cargo build --target aarch64-apple-ios
# run on iOS device
cd iOS
pod install
# then use Xcode to open fluid_demo.xcworkspace