At Mon, 17 Oct 2011 11:41:38 -0700,
Ian Lance Taylor wrote:
>
> Momchil Ivanov <momchil@xxxxxxx> writes:
>
> >> In gcc 4.5 and up -std=c99 implies -fexcess-precision=standard, q.v.
> >
> > Thank you for the tip, but I still don't understand why the following produces different output for both floats when compiled with -fexcess-precision=standard:
> >
> > #include <stdio.h>
> >
> > int
> > main(int argc, char *argv[])
> > {
> > float a = 268517138.f;
> > printf(" 268517138.f = %f\t%f\n", 268517138.f, a);
> >
> > return 0;
> > }
> >
> > output with -fexcess-precision=standard:
> > 268517138.f = 268517138.000000 268517152.000000
> > output without -fexcess-precision=standard:
> > 268517138.f = 268517152.000000 268517152.000000
>
> Good point. I think it is a bug that the values differ when using
> -fexcess-precision=standard. There is an implicit cast to double when
> calling printf; the problem may be that that cast is not being applied
> to the properly rounded value.
>
> Please open a bug report as per http://gcc.gnu.org/bugs . Thanks.
>
> Ian
bug report filed: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50773
