Final Thesis: Extending Cross-Platform Java Virtual Machine for Deployment in Automotive Platforms and Sustainable Software Production

Abstract: The automotive industry’s transition toward Software-Defined Vehicles demands middleware that executes across heterogeneous Electronic Control Units running incompatible operating systems. Despite Java’s “write once, run anywhere” promise, deploying Java-based middleware (like a network application) on QNX real-time systems remains challenged by the absence of open-source Java Virtual Machine support and strict determinism requirements. This thesis presents the design, implementation, and evaluation of OpenJDK 21 HotSpot ported to QNX Neutrino 8.0 on AArch64, enabling cross-platform Java middleware deployment without platform-specific reimplementation. The systematic porting methodology progresses through five interdependent phases: build system integration, toolchain adaptation, Operating System Abstraction Layer implementation, POSIX compatibility refinements, and Java standard library integration. Evaluation reveals fundamental architectural incompatibilities between modern garbage collection algorithms and microkernel memory semantics. G1GC achieves only 37.5% benchmark success versus 53.1% for SerialGC, establishing SerialGC as requisite for embedded stability. Performance characterization identifies a counter-intuitive inverse relationship between heap size and concurrency stability, driven by competition between Java heap and native thread stacks within QNX’s strictly-committed memory pool. The implementation provides a sustainable, open-source alternative to proprietary commercial JVMs, extending automotive middleware lifecycle across heterogeneous hardware generations while reducing development costs by eliminating redundant platform-specific porting effort.

Keywords: SDV, Software-Defined Vehicle, Java, QNX

PDF: Master Thesis

Reference: Kavipriyan Loganathan. Extending Cross-Platform Java Virtual Machine for Deployment in Automotive Platforms and Sustainable Software Production. Master Thesis. Friedrich-Alexander-Universität Erlangen-Nürnberg: 2025.