Clemens

@oscarbenjamin I don't think it's practical to build an alternative implementation that can rival numpy arrays. Arbitrary dimension arrays, slicing in all its generality, broadcasting, hundreds of existing functions for...

Personally, I would have a use for it in my baryrat package where it would be very useful for computing roots and poles of rational functions. Yes, it seems that...

During a quick search I didn't find *any* software package, no matter the language, that can solve generalized eigenvalue problems in arbitrary precision. It may exist somewhere out there, but...

> Is there a reason why you cannot reduce this to the standard eigenvalue problem (your M is not invertible?) Yes, M is singular in my application. Thanks for the...

I've dug into the performance question a bit and done some benchmarking. I noticed that arithmetic operations on gmpy real/complex numbers are across the board around 10x faster than operations...

> In my testing, I got similar results: Numpy array with mpmath content give ca. a 10x speed improvement for matrix operations That's surprising to hear for me. I'm quite...

> But on the other hand, using arb matrix operations directly give a further speedup, so the numpy question is kind of moot. I'd like to compare my results to...

Thanks! That did work for me, but I haven't gotten around to doing benchmarks using flint yet. I still want to do that soon. For now I've repackaged the linear...

That's great news. Thanks also for your last remark, I changed the attribution in flamp correspondingly. Here are the promised benchmarks including python-flint/arb. This is for computing the eigenvalues and...

Yes, I think it's very possible to integrate this back into mpmath. My first idea was to do it as a new context, but I don't know the mpmath internals...