Re: Changing the default i387 precision

[Matthias Drochner <M.Drochner%fz-juelich.de@localhost>]

On Sat, 4 May 2013 08:50:29 -0400
Greg Troxel <gdt%ir.bbn.com@localhost> wrote:
>    long double variables get the 80-bit precision
> 
> without
> 
>    making (non-long) doubles get results that don't comply with
> IEEE754 (by using extended precision intermedate values)

Here is a little test program which still fails on Linux/x86_64
(gcc 4.6.3) if compiled with -mfpmath=387 -O2.
So it is still the sad fact that you can't have full "long double"
support without compromising "double" accuracy. At least with gcc.
Would be interesting to see how clang does.

best regards
Matthias

#include <stdio.h>

#define FPTYPE double

void
test(FPTYPE x, FPTYPE y)
{
        const FPTYPE y2 = x + 1.0;
        if (y != y2)
                printf("error\n");
}

int
main()
{
        const FPTYPE x = .012;
        const FPTYPE y = x + 1.0;

        test(x, y);
        return 0;
}

Attachment: signature.asc
Description: PGP signature