This repo compare speed of
- taichi MLS-MPM
- houdini apic flip
- maya bifrost graph
- 32 x 32 x 32 grid resolution
- ~8000 particles
- 1m x 1m x 1m container
- 0.1m x 0.1m x 0.1m particle source at position (0, 0.5, 0)
- gravity is set to (0, -9.8, 0)
Measure result per substep:
- taichi cpu: ~2ms
- taichi gpu: ~0.28ms
- houdini: ~5ms
- maya: ~1000ms
hardware: CPU=AMD5900X, GPU=RTX2070
cost=2ms, substep=3:
- ngrid = 32, n_particle = 60000
- ngrid = 64, n_particle = 70000
- ngrid = 128, n_particle = 0
run:
python taichi_profiler.py