The Java Language Specification (JLS) has guided the evolution of one of the world’s most widely deployed programming languages for over two decades. Yet, despite Java’s robust concurrency libraries (e.g., java.util.concurrent ), the language still lacks first‑class syntax and semantics for expressing fine‑grained parallelism that modern heterogeneous hardware demands. This article proposes , a lightweight extension to the existing JLS that introduces native parallel‑programming constructs, a memory‑model‑aware type system, and a set of compiler‑runtime hooks designed to interoperate seamlessly with existing Java code‑bases. We present the design rationale, a concrete syntax proposal, a formal semantics sketch, and an early prototype implementation that demonstrates up to 3.2× speed‑up on typical data‑parallel workloads while preserving Java’s “write once, run anywhere” promise.
JLS PP does replace the JMM. Instead, it refines it: The Java Language Specification (JLS) has guided the
: Use a specialized utility to strip away legacy drivers that might cause conflicts with the JLSPP framework. We present the design rationale, a concrete syntax
| Benchmark | Baseline (Java SE 23) | JLS PP ( par for ) | Speed‑up | |-----------|----------------------|-------------------|----------| | Vector addition (10⁸ elements) | 1.84 s | 0.58 s | 3.2× | | Monte‑Carlo π estimation (10⁹ samples) | 6.12 s | 2.01 s | 3.0× | | Parallel quicksort (1 M integers) | 0.96 s | 0.33 s | 2.9× | | Benchmark | Baseline (Java SE 23) |