I’m happy to announce the release of Numba 0.50.0, which requires the new llvmlite 0.33.0.
Highlights of core feature changes include:
- The compilation chain is now based on LLVM 9.
- The error handling and reporting system has been improved to reduce the size of error messages, and also improve quality and specificity.
- The CUDA target has more stream constructors available and a new function for compiling to PTX without linking and loading the code to a device. Further, the macro-based system for describing CUDA threads and blocks has been replaced with standard typing and lowering implementations, for improved debugging and extensibility.
IMPORTANT: The backwards compatibility shim, that was present in 0.49.x to accommodate the refactoring of Numba’s internals, has been removed. If a module is imported from a moved location an
ImportError will occur.
Wheels are in the usual place, and conda packages are in the numba channel:
conda install -c numba numba=0.50.0
conda install -c numba llvmlite=0.33.0
(Packages should appear in the Anaconda repository in the next week or so.)
Please see the full changelog at:
And Stuart has made another great demo notebook for the new features you can try out in Binder:
Thanks again for everyone’s contributions and suggestions!