Re: [PATCH, gfortran testsuite]: Do not load denormals in gfortran.fortran_torture/intrinsic_nearest.f90

[IainS <developer at sandoe-acoustics dot co dot uk>]

On 23 Aug 2009, at 16:50, N.M. Maclaren wrote:

> On Aug 23 2009, Tim Prince wrote:
>
> > I ran into a peculiarity with gfortran C interoperability.  It  
> > throws a fatal KIND mismatch, as KIND(.false.) is 4, and KIND 
> > (c_bool) is 1.  I assumed that IEEE_SET_UNDERFLOW_MODE(gradual)  
> > should work with
> > logical(c_bool) :: gradual
> > Have I gone wrong in my assumption, or does the standard require
> > CALL IEEE_SET_UNDERFLOW_MODE(LOGICAL(.false.,KIND(C_BOOL)))
> > if IEEE_SET_UNDERFLOW is written according to iso_c_interop  ?
>
> Oh, gosh, you really do need a copy of the standard to work from.   
> IEEE
> support has nothing to do with C interoperability.   
> IEEE_SET_UNDERFLOW_MODE
> etc. are intrinsics, and follow whatever rules Fortran specifies - NOT
> necessarily ones that any normal procedure can follow - in this  
> case, its
> argument is specified to be a default LOGICAL scalar, so you should  
> be using
> just .FALSE.

I definitely need a copy of the standard to work from....

I especially need to understand what to do when the function cannot  
be honored on a specific target.

.. can you help with a link? or an appropriate abstract.

thanks
Iain