← Back to Monograph Index

🏎️ BIDSim vs. MCMC Drag-Race Benchmark

Official Metrological Drag-Race (Paper 4 / Benchmark 6.2 Settings)

πŸ“Š Benchmark Specs: 10 Candidates (100,000 Scenario Draws)

β€’ BIDSim WASM Engine (10-Core SIMD): 75.0 ms (Phase 1 Compilation: 64.1 ms, Phase 2 Extraction: 10.9 ms)

β€’ BIDSim Single-Threaded (Python/JS): 857.6 ms

β€’ Microcanonical HMC (MCHMC): 2,490.0 ms (31.6x Slower than BIDSim WASM)

β€’ Continuous MCMC (JAX-NUTS): 6,070.0 ms (80.9x Slower than BIDSim WASM)

⚑ BIDSim WASM Engine (Discrete Urns) 0.0 ms
🐒 Microcanonical HMC (MCHMC) 0.0 ms
🐌 Continuous MCMC (JAX-NUTS / NumPyro) 0.0 ms
--
Speedup vs JAX-NUTS
0.000
Autocorrelation (Lag-1)
100%
Effective Sample Size (ESS)
< 154 KB
Binary WASM Size
Code 2 : demo2-wasm-engine.html (Inférence Discrète BID Pure, PEMPI, Sans Dirichlet)