TradeCore Matching Engine

TradeCore is an institutional-grade, zero-allocation Matching Engine designed for ultra-low latency execution and strictly deterministic market replay. Built in modern C++20, it is specifically architected to avoid the OS heap, virtual dispatch, and cache-line tearing on the hot execution path.

🚀 Key Features

• Zero-Allocation Execution: The primary matching loop (ExecutionResult match(OrderAction)) exclusively uses pre-allocated memory pools and inline StaticVector containers, ensuring 0 heap allocations and 0 GC jitter post-startup.
• Deterministic Replay: The custom ReplayEngine ensures identical $O(1)$ topological reconstruction of historical ITCH 5.0 PCAP data across multiple playback modes (MaxThroughput, RealTime, Accelerated, SingleStep).
• Compile-Time Risk Pipeline: Risk policy evaluation completely avoids virtual vtable lookups by unrolling verification checks at compile-time via C++17 variadic fold expressions.
• Bounded Latency Telemetry: The tradecore::telemetry layer uses lock-free, bounds-checked HDR-style histograms to provide true sub-nanosecond precision performance mapping (Parser, Matcher, Risk, E2E).

🧠 Architectural Overview

For deep dives into the engine's design, refer to our comprehensive institutional documentation suite:

• Architecture & Pipeline Design
• Memory Layout & Cache Optimization
• Performance Guide & Tradeoffs
• Developer Onboarding Guide

⚡ Performance Contracts

TradeCore enforces statistical latency bounds mathematically via CI. We do not just report benchmarks; we fail builds that break these thresholds.

Component	Target Contract (P99)
ITCH Parsing	< 120 ns
Book Mutation	< 140 ns
Matching (Limit)	< 200 ns
Matching (Sweep)	< 250 ns

🛡️ Validation Matrix

To guarantee institutional-grade safety, TradeCore implements a layered topological verification matrix.

Verification Layer	Invocation	Scope	Pass Criteria
TRADECORE_VERIFY_BOOK	Debug / Tests	Asserts bidirectional AVL parent pointers, correct queue depths, and $O(1)$ BBO caches across the entire OrderBook topology.	Zero assertion failures
TRADECORE_VERIFY_MATCH	Per Execution Sweep	Validates size and price bijectivity during atomic limit sweeps. Guarantees price improvement belongs to the aggressor.	Zero assertion failures
TRADECORE_VERIFY_EXECUTION	Post Execution Pipeline	Macro-level check guaranteeing cumulative total executed volume plus residual volume exactly matches initial incoming volume.	Zero assertion failures
BookDigest	E2E Replay Validation	Deterministic FNV-1a state hash generated from an exact in-order tree traversal.	Successive runs yield identical digests

🛠️ Quick Start

git clone https://github.com/Zylus08/TradeCore.git
cd TradeCore
mkdir build && cd build
cmake -DCMAKE_BUILD_TYPE=Release ..
cmake --build . -j $(nproc)
ctest --output-on-failure

For detailed usage, see the Developer Guide.

📄 License & Contributing

• Please see CONTRIBUTING.md for pull request guidelines.
• Licensed under the MIT License. See LICENSE for details.