Performance

Optimization and performance considerations for cavity-coupled molecular dynamics.

GPU vs CPU

Performance comparison:

System Size	CPU (steps/s)	GPU (steps/s)	Speedup
512 molecules	500-2000	50,000-150,000	50-100×
2048 molecules	100-500	100,000-300,000	200-600×

Recommendation: Use GPU for N>500 molecules.

Timestep Optimization

Adaptive timestep:

from cavitymd.simulation import AdaptiveTimestepUpdater

adaptive = AdaptiveTimestepUpdater(
    initial_dt=0.001,
    min_dt=0.0001,
    max_dt=0.002,
    target_energy_drift=1e-4
)

sim.operations.updaters.append(adaptive)

Balances accuracy and speed automatically.

Scaling Studies

Strong scaling: Fixed system size, increase processors

Weak scaling: System size scales with processors

See HOOMD documentation for parallelization strategies.

Memory Optimization

GPU memory usage:

\[\text{Memory} \approx N_{\text{particles}} \times (48 \text{ bytes/particle})\]

Modern GPUs (8+ GB) handle 10,000+ molecules easily.

Back to: Cavity HOOMD User Guide