Technology news and discussion on TalkFuse.
TotalView adds Python debugging for high-performance computing workloads
Perforce has added Python support to TotalView, its parallel debugger for high-performance computing, allowing developers to inspect Python applications running across many processes. The change reflects Python’s growing use in scientific computing and distributed workloads, where debugging failures across thousands of parallel tasks can be difficult.
TotalView previously supported C/C++ and Fortran; its Python debugging uses Microsoft’s debugpy engine and can group hundreds of stack traces into one view. Developers can set breakpoints, inspect variables and evaluate Python expressions. It does not yet provide full cross-language debugging for Python code calling C/C++ libraries, though Perforce plans to combine those workflows in a future release.
In short
- TotalView now supports debugging parallel Python applications.
- It can consolidate hundreds of stack traces.
- Cross-language Python and C/C++ debugging is still planned.
Explained simply
High-performance computing uses powerful computers to split demanding tasks across many processors, which run parts of a job at the same time. Scientists use this approach for work such as modelling, simulations and analysing large datasets.
TotalView is a debugging tool from Perforce, a software company. It has supported C, C++ and Fortran, languages often used in high-performance computing; Python is also increasingly used for scientific and distributed applications.
When a job runs across many processes, a failure can be difficult to trace because each process may have its own sequence of operations. Python debugging in TotalView lets developers inspect those processes together, while mixed Python and C or C++ code remains a separate challenge.
Read the full article at the source →
Originally published by The Register as “Python parallel debugging comes to HPC thanks to TotalView upgrade”.