Closed Foggalong closed 4 months ago
Looking forward, it may be hard to continue using Python for this, as the linearization-based code - which is effectively sequential quadratic programming (SQP) as a way of solving a nonlinear programming problem (NLP) - is going to be algorithmic. When working with very large problems, a Python code would have to use utilities to do the algorithmic "set-up", since writing it in pure Python would be a huge performance hit.
These utilities - like the HiGHS QP solver you'd be using - aren't written in Python. The "real" scientific computing is done with code written in C++, C or Fortran
@jajhall Yeah most all of Python's "under the hood" work isn't in Python itself, even basic NumPy is mostly written in C. For all of this stuff, Python is mostly just the interface through which it's manipulated.
Merge not intended, this is just so I can reply to Julian's code comments