Top picks for Python readers on InfoWorld
Making Python faster won’t be easy, but it’ll be worth it
Python’s malleability and dynamism make it powerful, but also very hard to speed up. Here’s a look at some of the ways developers are working to make Python faster from within.
Understand Python’s new lock file format
Lock files for project dependencies have long been missing from Python, but that’s about to change. Here’s what you need to know about using the new common lock file format in your Python projects.
The power of Python’s editable package installations
Want to unlock a new superpower for working with Python packages? Here’s how to install a package, edit changes to the source, and have the changes show up automatically wherever it’s installed.
Exploring new features in Cython 3.1
The next version of the Python-to-C compilation tool isn’t quite out yet, but you can get a head start on all its powerful new features, including support for Python’s free-threaded/“no-GIL” build.
More good reads and Python updates elsewhere
Is Python code sensitive to CPU caching?
You might think Python’s too high-level for CPU caches to matter, but Lukas Atkinson’s research reveals something different.
NVIDIA adds native Python support to CUDA with cuda.core
This latest addition to NVIDIA’s CUDA toolkit lets you plug straight into CUDA on supported systems with a single pip install. (Also see the discussion at The New Stack.)
How to run PyTorch on the free-threaded builds of Python
Here’s how to achieve faster inference for PyTorch models—and by an order of magnitude in this case! It’s all thanks to Python’s new “no-GIL” build.
Just for fun: How C code becomes assembly
It isn’t always obvious how C maps to assembly, but this deep dive into the Clang/LLVM compiler system will take you there.
