ahrav/performance-analyzer

Performance Analyzer

Analyze performance-critical code in the gossip-rs workspace (coordination, scanner engine, data structures).

When to Use

• After writing hot-path code (coordination, scanning, decoding, validation)
• Before committing changes to crates/scanner-engine/src/engine/ modules
• When benchmark results show unexpected regressions
• During optimization work on data structures in crates/gossip-stdx/src/

Analysis Checklist

Memory & Allocation

• [ ] Unnecessary allocations in loops (Vec, String, Box)
• [ ] Missing with_capacity() for known-size collections
• [ ] Cloning where borrowing would suffice
• [ ] Large structs passed by value instead of reference

CPU & Cache

• [ ] False sharing in concurrent data structures
• [ ] Cache-unfriendly access patterns (strided, random)
• [ ] Missing #[inline(always)] on hot functions
• [ ] Branch-heavy code that could use branchless alternatives

Async & Concurrency

• [ ] Blocking operations in async contexts
• [ ] Lock contention patterns
• [ ] Oversized futures (check with std::mem::size_of)
• [ ] Unnecessary Arc when Rc or ownership would work

Rust-Specific

• [ ] Bounds checks in hot loops (consider get_unchecked with proof)
• [ ] Iterator vs manual loop performance
• [ ] &str vs String in function signatures
• [ ] Zero-cost abstraction violations

Project-Specific Patterns

This codebase uses these performance patterns:

• NONE_U32 = u32::MAX as sentinel (avoid Option overhead)
• #[inline(always)] on hot-path functions
• debug_assert! for invariant checks (zero cost in release)
• Const generics (G) for compile-time granularity selection

Output Format

## Performance Analysis: [file/function]

### Findings

| Severity | Issue | Location | Impact |
|----------|-------|----------|--------|
| HIGH | Allocation in hot loop | line:XX | ~Xns per call |
| MEDIUM | Missing inline hint | line:XX | Potential call overhead |

### Recommendations

1. **[Issue]**: [Specific fix with code example]
   ```rust
   // Before
   // After

Validation

Run these benchmarks to verify:

• cargo bench --bench <relevant_bench>

## Related Skills

- `/bench-compare` - Before/after measurement
- `/asm-forge` - ASM-guided optimization
- `/perf-regression` - Full regression workflow