Sergey B Kirpichev

Hmm, probably you are right, maybe it worth to keep this separate. Both are about automatic tuning of parameters for this specific solver. On another hand, I appreciate more info:...

Attached code could reproduce issue using mpmath-only functions: [a.py.txt](https://github.com/mpmath/mpmath/files/14385774/a.py.txt) I would appreciate a simpler variant, however;) To me this looks like an enhancement requests, not a bug. Unless you have...

> I don't think most people know about functools.reduce You can do same, using simple for loop: ``` In [1]: from mpmath import * In [2]: a = mp.matrix([[1, 2],...

Thanks for your contribution and patience, but I'm closing this pr. As it mentioned above, the current version has nothing specific to mpmath c.f. the ``np.kron()``, you could just use...

fp part was fixed in #798

```pycon >>> iv.eig(fp.matrix([[1,2], [3,4]])) Traceback (most recent call last): File "", line 1, in File "/home/sk/src/mpmath/mpmath/matrices/eigen.py", line 761, in eig hessenberg_qr(ctx, A, Q) File "/home/sk/src/mpmath/mpmath/matrices/eigen.py", line 481, in hessenberg_qr qr_step(ctx,...

@warnerjon12, the problem is that rich-comparison helpers for iv sometimes return ``None``. In a context, where boolean value is required (e.g. in min/max) - Python calls ``bool(None)``, that's ``False``.

@sven-mayer, not sure how you run above benchmarks (perhaps, with pygsl). With ctypes I got 1000x difference: ``` $ python -m timeit -u usec -s 'from ctypes import CDLL, c_double;l=CDLL("libgsl.so");l.gsl_sf_clausen.restype...

> I'm currently reading about Arb and FLINT available in python via python-flint Yeah, I think it should be available in Arb. Not sure if the FFT support is available...

Perhaps, this plugin should use AST instead. Then ``get_docstring_tokens()`` helper will be trivial.