Matching Engine

matchingimplemented by → rsx-matching← TopologyAll Components

What it is

One ME instance runs per symbol, single-threaded and pinned to a dedicated core. It uses a slab arena allocator (pre-allocated OrderSlots) and a CompressionMap for O(1) level access. Price-time FIFO matching with GTC/IOC/FOK. Measured at 54 ns per fill and ~340 ns algorithmic round-trip.

Live Health (status, RSS, details)

CPU omitted — busy-spin tiles peg 100% by design (healthy).

loading...

Latency

match single fill
54 ns
Criterion bench
algo round-trip
~340 ns
component bench
insert order
857 ns
Criterion bench

These are Criterion micro-benchmarks. No per-component live latency from the running system — see Latency tab for the end-to-end probe.

Log tail (matching)

loading...

Live Orderbook (symbol 10 / PENGU)

loading...