Re: [boost] Formal Review for Interval Library(was:IntervalLibraryreminder)

From: Joerg Walter (jhr.walter_at_[hidden])
Date: 2002-09-08 03:27:15

----- Original Message -----
From: "Gabriel Dos Reis" <gdr_at_[hidden]>
To: <boost_at_[hidden]>
Sent: Sunday, September 08, 2002 9:46 AM
Subject: Re: [boost] Formal Review for Interval
Library(was:IntervalLibraryreminder)

> jhr.walter_at_[hidden] (Joerg Walter) writes:
>
> | > Well, happy or not one has to cope with one thing or the other.
> | > A complaint I regularly see reported on the GCC list is GCC's
> | > inconsistent use of mixed precision on x86 plateforms.
> |
> | But wouldn't help an implementation like
> |
> | template<class T1, class T2>
> | typename promote_traits<T1, T2>::result_type operator+(const T1 &t1,
const
> | T2 &t2) {
> |   typedef  typename promote_traits<T1, T2>::result_type  result_type;
> |   return result_type (t1) + result_type (t2);
> | }
> |
> | to prevent such complaints then?
>
> No.  Sure the computation is done in a wider precision, but there is no
> control on the *actual* destination type.

I see. At this point we usually need special functions, to compute the
double precision inner product for example.

Thanks

Joerg