
Perforce Software has introduced parallel Python debugging capabilities to its TotalView development tool, allowing engineers to investigate problems across multiple processes and computing nodes simultaneously.
The update, announced on 17 November 2026, is designed for developers working with high-performance computing (HPC), artificial intelligence (AI), machine learning and scientific simulations, where tracking down software failures can be particularly complex and expensive.
Python has become an increasingly important language for AI development. According to GitHub’s Octoverse 2025 report, nearly half of new AI projects on GitHub were built primarily in Python.
However, debugging Python applications running across distributed computing environments presents challenges that are not typically encountered when working with individual scripts.
Failures can occur across multiple processes or computing nodes, making it difficult for developers to identify their source. In large-scale AI and scientific computing environments, prolonged debugging can also consume valuable processing capacity and delay development.
Addressing the challenges of distributed Python
The new TotalView capability allows developers to view and control multiple Python processes at the same time, rather than investigating each one individually. It supports distributed Message Passing Interface (MPI) workloads and is intended to help engineers identify faults that may occur across different processes and nodes.
Perforce developed the functionality following discussions with teams at Lawrence Livermore, Sandia and Los Alamos national laboratories to better understand the debugging requirements of scientific computing and machine learning environments.
The company says the update will also benefit teams working with applications that combine Python and other programming languages, including C++.
Developers already using TotalView for MPI, OpenMP, C/C++ and Fortran applications can now extend their existing debugging workflows to parallel Python without introducing a separate specialist tool.
Jeff Michael, Senior Director of Product Management at Perforce, said:
“Teams are building more of their most important HPC, simulation, and AI work in Python, and the compute they run it on is too valuable to spend on slow debugging.”
He added that the new functionality extends TotalView’s established HPC debugging capabilities to parallel Python, helping teams identify problems earlier and reduce the resources consumed during development.
The cost of debugging AI workloads
As AI models and scientific applications become more computationally demanding, the resources required to develop and test them are an increasing consideration for engineering teams.
Distributed applications can involve numerous processes operating simultaneously, with failures sometimes occurring in only one part of a much larger workload.
Identifying the source of those failures can require extensive investigation, particularly when developers have limited visibility across the entire application.
Bill Burns, VP of Software Engineering at Perforce, explained:
“Debugging a parallel Python application is very different from debugging a single script.”
“When a failure occurs on one rank among many, developers need to see what every Python process is doing and control them together.”
The new functionality is intended to address that problem by providing a consolidated view of processes across distributed environments, allowing developers to investigate failures without examining each process separately.
For organisations running computationally intensive AI and simulation workloads, reducing the time spent identifying software faults could help improve development efficiency and limit unnecessary compute usage.
Availability
Parallel Python Debugging is now generally available in Perforce TotalView.
Perforce is demonstrating the capability at SC26, taking place from 17–19 November 2026, and will host a technical webinar on parallel Python debugging on 2 December.
Further information is available on the [Perforce TotalView website](https://www.perforce.com/reso


