MAR 28, 2026
XCX 2.2: PERFORMANCE & EFFICIENCY
CORE PERFORMANCE
- JIT Compilation — Cranelift-based JIT; 1B-iteration triple nested loop benchmark: ~478ms vs ~140s in 2.1 (~293x speedup).
- Trace Recording — Automatic trace capture and native code generation for loops.
- NaN-Boxed Values — Optimized 64-bit value representation (8 bytes).
- Lock-Free Atomics — Improved concurrency with AtomicBool shutdown handling.
VM IMPROVEMENTS
- Optimized Opcodes — New IncLocalLoopNext, IncVarLoopNext, GuardInt/Float for faster execution.
- Better Fiber Support — Enhanced FiberState with yield value caching.
MIGRATION NOTES
| Feature | 2.1 | 2.2 |
|---|---|---|
| Execution | Interpreted | JIT + Interpreted |
| Value Size | 24+ bytes | 8 bytes |
| Loop Perf (1B) | ~140s | ~478ms |
IMPORTANT Bytecode is not compatible with 2.1. Requires cranelift, parking_lot, and argon2.