I’m a PhD researcher in the Programming Languages & Systems group at the University of Kent, supervised by Dr Stefan Marr.
My research focuses on a long-standing challenge: why Java profilers are so inaccurate, and why they disagree on what’s “hot.”
Sampling profilers often mislead developers, while traditional instrumentation can distort performance by orders of magnitude.

I’m developing a hybrid profiler for the Graal compiler that:

• 🧩 Inserts lightweight probes after all optimisation phases — leaving the compiler’s behaviour unchanged.
• ⚙️ Combines sampling to locate hot code with instrumentation to zoom in precisely.
• 🧠 Enables fine-grained, compiler-level slowdown for accurate ground-truth evaluation.

Beyond building better profilers, I’m also designing techniques to measure profiler accuracy itself, introducing controlled slowdowns that reveal whether profilers can detect the true performance behavior.

🎯 Goal: Produce accurate profiles with minimal overhead — and a reliable way to prove that accuracy.

I’m currently building an instrumentation testbed for the Graal compiler to explore:

• How sampling and instrumentation interfere with program execution.
• How compiler-level instrumentation can be made lightweight and predictable.
• How to quantify profiler accuracy across workloads and runtimes.

• Burchell, H., & Marr, S. (2025).
  Divining Profiler Accuracy: An Approach to Approximate Profiler Accuracy Through Machine Code-Level Slowdown.
  Proceedings of the ACM on Programming Languages (OOPSLA ’25).

• Burchell, H., & Marr, S. (2025).
  Evaluating Candidate Instructions for Reliable Program Slowdown at the Compiler Level: Towards Supporting Fine-Grained Slowdown for Advanced Developer Tooling.
  17th ACM SIGPLAN International Workshop on Virtual Machines and Intermediate Languages (VMIL ’25), Singapore.

• Burchell, H., Larose, O., & Marr, S. (2024).
  Towards Realistic Results for Instrumentation-Based Profilers for JIT-Compiled Systems.
  21st ACM SIGPLAN International Conference on Managed Programming Languages and Runtimes (MPLR ’24), Vienna.

➡️ View all publications in KAR

If you’re interested in:

• Profilers and instrumentation
• CPU sampling and performance analysis
• Compiler-level slowdown and benchmarking

Feel free to get in touch!

📫 Email: [email protected]

_{© 2025 Humphrey Burchell — Researching how to make profilers both accurate and trustworthy.}