Other Python accelerators: JAX , pyccel , pystencils …

Good target for Transonic and Pythran: https://github.com/ogrisel/pygbm

For Pythran, we would need more OOP support.

Good Cython and Numba examples for Transonic:

• https://github.com/jeremiedbb/tutorial-euroscipy-2019

• http://pycodegen.pages.walberla.net/pystencils/notebooks/demo_benchmark.html

We need something like a NumFocus Europe.

For Cython, we have to fully work with fused types

For Cython, we need to be able to specify if an array will be used as a memoryview or a np.ndarray .

Transonic should soon support A = Array[int, "1d", "memview"]

We have to work more on type annotation of Numpy code, see numpy-stub .