Parallel Python Debugging in HPC: Challenges, Approaches, and Real-World Lessons

Python is now a first-class language in high-performance computing (HPC), powering simulations, AI/ML pipelines, and data-intensive workflows at national laboratories and research institutions. But traditional single-process debuggers were not built for workloads spanning thousands of MPI ranks, GPU accelerators, or hybrid Python/C++ architectures. 

Join Perforce VP of Software Engineering Bill Burns and Perforce Senior Manager of Software Engineering Jasmit Singh on December 9^th for a practical look at the challenges of debugging parallel Python applications and the capabilities development teams need to diagnose failures at scale. Drawing on collaboration with teams at Lawrence Livermore, Sandia, and Los Alamos national laboratories, this webinar will explore real-world requirements and lessons from distributed scientific and machine learning workloads. 

What You’ll Learn 

• Why conventional Python debuggers fall short for distributed HPC workloads. 
• How coordinated multi-process debugging helps teams control and inspect Python interpreters across ranks, nodes, and roles. 
• What capabilities are needed to compare distributed state, identify outliers, and isolate scale-only failures. 
• How evolving Python concurrency, including GIL-less execution, changes debugger requirements. 
• Lessons from real-world scientific computing and ML data pipeline use cases at DOE laboratories. 
• And more!