Not that I can think of at the moment.
Sometimes these bugs can be terribly subtle. The other day I tracked
down a bug in FLINT's F_mpz_div_2exp on 32 bit machines. In the end I
found the following code gave the wrong answer:
ulong a = 573498595893479UL;
ulong b = 42;
printf("%lu\n", a>>b);
The problem is, on a 32 bit machine 42 >= 32 and the shift operator
does not work on some architectures when that is the case. For the
longest time I thought this was a compiler bug!!
Bill.
On Jan 29, 11:48 am, Sebastian Pancratz <s...@pancratz.org> wrote:
> On Jan 29, 10:13 am, Bill Hart <goodwillh...@googlemail.com> wrote:
>
> > I wonder if it is possible to recreate the problem just using FLINT,
> > without any Sage, i.e. just write a short FLINT program which
> > replicates the problem.
>
> > I don't understand how the multiplication could take any serious
> > quantity of time unless the pow was screwed up somehow. But perhaps it
> > is.
>
> > Could you insert some traces to print the values of m, the number of
> > limbs of lead and (when it is computed) the number of limbs of t.
>
> > Bill.
>
> Dear Bill,
>
> At least using a naive approach of simply writing a short C program to
> go through the same FLINT calls with the same mathematical objects
> did, unsurprisingly, not reproduce the problem. However, I did not
> ensure that all objects had the same state (length of the coefficient
> array of polynomials, limbs for the coefficients, limbs of all other
> fmpz_t's involved). Other than the above three pieces of data, which
> I can print out between the FLINT calls in the code for every
> fmpz_poly_t and fmpz_t involved --- I'll do this right after a seminar
> which begins in a few minutes ---, is there any other piece of
> information that is relevant to representing the state?
>
> Thanks,
>
> Sebastian
--
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org
