Why is GraalVM's JIT compiler often faster than the default HotSpot C2 compiler?

Graal tends to outperform C2 on certain workloads because of differences in its intermediate representation and optimization pipeline, not because it's simply "newer."

Graal uses a unified sea-of-nodes graph IR that represents both control flow and data flow together, which makes global optimizations - like more aggressive method inlining across call sites, partial escape analysis that can scalar-replace objects even along some but not all control-flow paths, and speculative optimizations backed by deoptimization - easier to implement and compose than in C2's more rigid pipeline.

Because it's written in Java rather than C++, Graal's optimization passes are also easier for Oracle and the community to extend and tune, meaning new optimizations land faster.

That said, "often faster" isn't "always faster" - for some workloads, especially very short-lived ones, C2's decades of low-level tuning still wins, and results should be measured per-application rather than assumed.