On Thu, 2023-03-23 at 22:05 +0000, Dima Pasechnik wrote:
> On Thu, Mar 23, 2023 at 9:14 PM Michael Orlitzky <
mic...@orlitzky.com> wrote:
> >
> > Does someone have access to
https://github.com/sagemath/FlintQS ?
> >
> > The build is broken with clang-16 and there's an open pull request to
> > fix it. A new release would be helpful.
>
> I promoted you to a maintainer of this repo.
>
Thanks. After spending some time with the code, I'd rather replace it.
There are unreported security issues with its use of /tmp,
and it's just generally pretty fragile. It also looks like a more
reliable (but often slower) implementation has been added to flint in
the meantime, as qsieve_factor().
On the examples I've tried, FlintQS is usually capable of factoring a
product of two primes, and relatively quickly. Switching the
implementation to qsieve_factor() slows it down a bit (2s to 3s, for
example). However, in cases like
qsieve(329884789374968270239373010571299817801707558079749117449)
which is a product of THREE primes, FlintQS gives you the wrong answer
in four seconds whereas qsieve_factor() takes six to get it right.
Number theorists, please speak up if you prefer to get less reliable
answers more quickly in some scenario that I haven't imagined.