Nuno Lopes

Contact E-mail: nuno.lopes@tecnico.ulisboa.pt Mail: INESC-ID, Office 134 Rua Alves Redol, 9 1000-029 Lisbon, Portugal Associate Professor @ Instituto Superior Técnico (Universidade de Lisboa) / INESC-ID GitHub / Linkedin

Funding

Our research has been generously supported by the following sponsors:

• Google (€35k)
• Woven Alpha ($30k)

I thank the continuous support of all of our sponsors. Our work wouldn't be possible without their support.
If you and/or company has benefitted from our work, for example, by using our free tools, our papers and presentations, or by hiring our former students, please consider sponsoring our research lab.

Publications

Last 5:

• An SMT Encoding of LLVM's Memory Model for Bounded Translation Validation (CAV'21)
• Supercharging Plant Configurations Using Z3 (CPAIOR'21)
• Alive2: Bounded Translation Validation for LLVM (PLDI'21, Distinguished Paper Award)
• AliveInLean: A Verified LLVM Peephole Optimization Verifier (CAV'19)
• Fast BGP Simulation of Large Datacenters (VMCAI'19)

Complete Publications List (DBLP, Scholar)

Slides

Current Projects

• Alive2 (Dashboard of pending LLVM issues; Online Alive2): Automatic verification of LLVM optimizations
• Torchy: A tracing JIT compiler for PyTorch

Students

I'm looking for great students that want to do Masters or PhD with me. Please get in touch if you are interested. Note that for Masters you must be enrolled with IST before contacting me.
Possible topics (non-exhaustive list):

• Compilers: IR design, scalable link-time optimizations, PGO, UB-aware optimizations
• Automatic software verification, namely of compilers (LLVM, MLIR, V8)
• Compilers and runtimes for machine learning frameworks like PyTorch or TensorFlow
• Code gen of machine learning kernels for many-core chips
• Law and regulations: formal specifications, model checking, optimization
• Security: automatic exploit generation, miscompilation exploitation
• Physical security: probablistic model checking
• Application of the Chinese Remainder theorem to SMT solving
• Blockchain mining: market efficiency (CPU/RAM/storage/network)
• Blockchain security: exploiting AMMs, compilers
• Trading/MM: volatily, cross-asset hedging, online ads

Former Students

PhD

• Juneyoung Lee, Seoul National University, 2021. A validated semantics for LLVM IR. Awards: student research competition (SRC) award at EuroLLVM'19, best PhD thesis award from the college of engineering of SNU.

MSc

• Sofia Salgueiro, IST-UL, 2021. Automatic synthesis of optimal datacenter networks.
• Kevin de Vos, IST-UL, 2020. Translation validation for the LLVM compiler.
• Filipe Azevedo, IST-UL, 2020. Evaluating value-wise poison values for the LLVM compiler.
• Janek van Oirschot, Eindhoven University of Technology, 2019. Extending tree pattern matching for application to peephole optimizations.

Past Projects

• Alive: Automatic verification of LLVM peephole optimizations
• CORK: Compiler Optimization Correctness Checking
• HSF: An automatic software verifier based on Horn Clauses
• PSyCO: Precondition Synthesis for Compiler Optimizations