Ilhan Polat

Before we get too excited about the array API parts and digress, please let's first identify if nanobind is a drop-in replacement for pybind11 or not. Otherwise, this is dead...

Thank you @wjakob for chiming and in general thank you for these wonderful tools. I guess, as you rightly point out, we first need to try it out with some...

Today I also learned that pybind11 will not be committed to stable API support but nanobind is/will. If I did not misunderstand this issue here https://github.com/pybind/pybind11/issues/1755, https://github.com/pybind/pybind11/issues/5652, and also @da-woods'...

Is there any discussion to read about for NumPy not supporting it? Are we using non-stable parts from CPython necessarily for NumPy C-API?

Man, this is such a nerd-sniping subject get lost in, but I'll try to resist. `f2py` can be avoided after fortran translations, Pythran has very little codebase and can be...

Not sure how to help but great that this is happening. The new C modules are pretty much in the "find/replace `int` -> `ILP_INT`" state. But the calls to BLAS/LAPACK,...

That's a very common Python dict C struct mapping. So if there is a problem it should error out but I think it might be that LongLong is not necessarily...

> Pretty much all Fortran->C rewrites are not ILP64-ready I did not want to barge in to the discussion so I'm waiting for this to stabilize and I'll start the...

Regarding the Cython based BLAS/LAPACK we learned to mimic f2py by stride tricks and silent F-major conversions etc. in our linalg work. Hence we can follow the argument sanitizing ->...

@ev-br I have some questions, I'll also reply your mail later. So this `_fitpackmodule.c` imports three functions from FITPACK and exposes a module `_fitpack`. But at the same time `dfitpack.pyf`...