[m-users.] Too Slow

[Mark Brown]

Hi Matthias,

Thanks for starting the ball rolling...

On Fri, May 1, 2015 at 10:02 PM, Matthias Guedemann
<matthias.guedemann at googlemail.com> wrote:
> Hi,
>
> here is a version with some basic error handling, i.e., failing with a
> message if either creating an mp_int fails or an operation does not
> return MP_OKAY.
>
> Any information about problems, bugs etc. is welcome.

> :- pred mp_init(int::in, mp_result_type::out, mp_result_type::out,
>               mp_int::out) is det.
> :- pragma foreign_proc("C",
>            mp_init(Value::in, InitResult::out, OpResult::out,
>                     Mp_Int::out),
>             [will_not_call_mercury, promise_pure, thread_safe],
>             "
>             int initResult, opResult;
>             Mp_Int     = MR_GC_NEW_ATTRIB(mp_int, MR_ALLOC_ID);
>             initResult = mp_init(Mp_Int);
>             opResult   = mp_set_int(Mp_Int, Value);
>             InitResult = initResult;
>             OpResult   = opResult;
>             ").

You call mp_set_int before checking the result from mp_init, which
could be a problem.

Rather than return two foreign results it may be nicer to define a
Mercury enumeration of the possible errors and export it with
foreign_export_enum (sections 14.5, 14.10.1.2), then create the
appropriate Mercury value from the C code.

Cheers,
Mark.