[PATCH] Document arithmetic overflow semantics

[Fergus Henderson]

On 13-Feb-2003, Robert Dewar <dewar@gnat.com> wrote:
> > Suppose we denote the three variants of "+" as follows:
> > 
> >       +undef          undefined behaviour on overflow
> >       +wrap           wraps on overflow
> >       +trap           traps on overflow
> 
> I still don't see how +wrap differs from an unsigned addition, so it seems
> simply confusing to have it around explicitly.

You are right that +wrap has the same semantics as an unsigned addition.
But I don't think it would be confusing to have it around explicitly,
and I do think that doing this would make things a lot easier.

In GCC trees the signedness of additions is determined by the signedness
of the types of the operands.  Representing +wrap additions as unsigned
operations would require that the operations be converted to unsigned
types.  This will require more tree nodes, which may have a compilation
time impact, and the additional conversions would complicate the intermediate
representation in ways that may inhibit some optimizations.

I think it would be much better to adopt a consistent approach which will
work for all arithmetic operations, not just plus and minus.

-- 
Fergus Henderson <fjh@cs.mu.oz.au>  |  "I have always known that the pursuit
The University of Melbourne         |  of excellence is a lethal habit"
WWW: <http://www.cs.mu.oz.au/~fjh>  |     -- the last words of T. S. Garp.