Re: [gfortran] Fix NULL reference types.

[Mark Mitchell <mark at codesourcery dot com>]

Daniel Jacobowitz wrote:

> On Thu, Aug 19, 2004 at 01:54:53PM -0400, Richard Kenner wrote:
>  
>
> >    I suppose that we could provide the same information I wanted to convey 
> >    with REFERENCE_TYPE to the optimizer in some other way: a bit on 
> >    VAR_DECLs, PARM_DECLs, FIELD_DECLs, etc. that says "never NULL".
> >
> > And INDIRECT_REF, which is somewhat of the harder one.  That's why I
> > think types are the best approach.
> >
> >    I do think that we should find some way to give this information to the 
> >    optimizers.
> >
> > The other issue is when we know that an indirect reference can't trap, such
> > as the CHAIN reference.
> >    
>
> I suspect we know that Fortran's reference types can't trap if they are
> non-NULL, also.  That might be nice to represent also.
>  
This is true in C++, too: a read/write through a reference will never 
trap in a valid C++ program.

Daniel's suggestion about range-information is pleasingly general, as 
long as the optimizers know that a pointer being "not 0" is a really 
special case -- much more interesting that it being "not 0xafeaf3ac" for 
example.

--
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com