[bonnhatnguyen]

Description

Adds the first C++ implementation (1.cpp) for the lru (Least Recently Used) cache benchmark, enabling C++ on the leaderboard.

Key Architectural Highlights

• Flat-Array Intrusive Doubly-Linked List + Hash Buckets: Uses contiguous flat vectors with 32-bit indices (nodes and buckets) rather than pointer chasing or separate allocations per node.
• Zero Allocations During Benchmark Loop: Memory for the maximum required capacity is preallocated upfront. All put and get operations run with zero heap allocations.
• L1-Cache Resident: With capacity $N \le 1,000$, the entire cache metadata fits comfortably inside L1 data cache (< 32 KB), ensuring near-instant cache hits and maximum memory bandwidth efficiency.
• Fibonacci Multiplicative Hashing: Power-of-two mask indexing with (key * 2654435761u) & mask providing excellent distribution and minimal collision chaining.

Verification & Benchmark

• Verified against all unit tests:
  • 10 1000 -> 10_1000_out (Match OK)
  • 77 7777 -> 77_7777_out (Match OK)
  • 100 10000 -> 100_10000_out (Match OK)
• Benchmark runtime on largest workload (1000 3000000 - 3,000,000 operations):
  • ~0.07s user time (~42.8 million operations/second)