Just ran through the benchmarks. The key insight isn't just about raw throughput it's about memory locality and GC pressure. When you pack primitives directly into the table instead of boxing to objects, you eliminate an entire layer of indirection and the cache misses that come with it. The numbers...
question