Bell Labs internship on Hot-Swappable JIT Compilation for Faster Cold Starts (PhD)

Beschreibung:

Modern runtimes (JavaScript, WebAssembly) rely on multi-tier JITs to combine fast startup with optimized performance. However, cold-start compilation and repeated re-optimization still incur significant cost overhead, especially in browsers, serverless platforms, and mobile environments.
This internship explores hot-swappable JIT-compiled code sections: externally pre-optimized fragments (e.g. on the server) that can be safely injected into an active interpreter or baseline JIT (e.g. in the browser). The goal is to reduce redundant compilation, accelerate cold starts, and evaluate tradeoffs between de-optimization and execution behavior.
A key challenge is how to maintain the security and sandboxing guarantees of these environments. Hence, we envision a need to define secure patchpoints, security mechanisms, and interfaces between the interpreter, baseline JIT, and injected optimized code.
A prototype can target JavaScript or WebAssembly (or others), using custom interpreters or existing open-source JIT infrastructures.
Duration: flexible, to be agreed (typically 3-5 months), starting time flexible.

* Build upon an existing (open source) JavaScript or WebAssembly runtime to pre-optimize "likely hot" code fragments.
* Explore mechanisms to maintain safety and sandboxing guarantees.
* Evaluate the trade-offs in performance gain (due to the optimization), proving time (for asserting correctness ahead of time), and verification time (for verifying safety during execution).
* Apply your research to a practical problem.
* Potential for a scientific publication with Bell Labs researchers

Ihr Profil:

* Student enrolled in Ph.D. Computer Science/Engineering.
* Solid background in compilers, JITs, interpreters, optimizers, or runtime systems.
* Strong programming skills, preferably in C, C++, or Rust.
* Language skills: English