Skip to content

Order Book Benchmark Results

These benchmark results compare the performance of the two order book implementations in Flox:

  • FullOrderBook: a general-purpose order book based on std::map
  • WindowedOrderBook: a performance-optimized order book with fixed-size price windows and ring buffers

All benchmarks were run on:

  • CPU: 8 x 4200 MHz
  • Cache: L1d 48 KiB, L1i 32 KiB (x4), L2 1280 KiB (x4), L3 8192 KiB (x1)
  • Load average: ~1.6
  • CPU scaling: enabled

⚠️ Note: CPU scaling may affect measurements. Real-time results may incur extra overhead.

FullOrderBook

-------------------------------------------------------------
Benchmark                   Time             CPU   Iterations
-------------------------------------------------------------
BM_ApplyBookUpdate        319 us          319 us         2045
BM_BestBid               7.96 ns         7.96 ns     87824155
BM_BestAsk               8.29 ns         8.29 ns     83782336

WindowedOrderBook

-------------------------------------------------------------
Benchmark                   Time             CPU   Iterations
-------------------------------------------------------------
BM_ApplyBookUpdate        200 us          200 us         3405
BM_BestBid               8.44 ns         8.44 ns     82694003
BM_BestAsk               7.74 ns         7.74 ns     82792804