What problem does it solve? C++ code in trading engines, indicators, and replay pipelines often ships with hidden performance problems—O(n^2) loops, unnecessary copies, wrong containers, and virtual calls in hot paths—that only surface as slow backtests or regressed benchmarks. This Skill gives reviewers and authors a concrete checklist to find and fix those smells with measurement, not guesswork. ## Core Features & Use Cases - Smell detection checklists: Tables of concrete anti-patterns for containers, copies/allocations, virtual calls, algorithms, I/O, strings, synchronization, and memory layout, each paired with a specific fix. - Measure-first workflow: Enforces a measure → fix → re-measure discipline, with profiling guidance (perf, Instruments, WPA/VTune) and rules for documenting reproducible before/after benchmarks. - Repository-specific guardrails: Ties optimization work to the project's hot paths (engine bar loop, indicator update(), replay, data prefetch) and forbids unmeasured optimizations, premature SIMD, and unjustified caches. - Use Case: While reviewing a pull request that adds a std::map lookup inside the engine's per-bar loop, use this Skill to flag the container choice, suggest a measured alternative, and require benchmark evidence before merge. ## Quick Start Ask the AI to review a C++ file or diff for performance problems in hot paths using the cpp-performance checklist.
