[m-dev.] .NET back-end test results

[Fergus Henderson]

On 06-Nov-2002, Peter Ross <pro at missioncriticalit.com> wrote:
> fjh wrote:
> > On 06-Nov-2002, Peter Ross <pro at missioncriticalit.com> wrote:
> > >
> > > This is the start of my fix for the problem B3 (FAILED.UNDEFINED)
> > > class by trying to expand out the equivalence types for the special
> > > predicates when producing the special predicates.
> >
> > I was thinking of doing this expansion at a later stage of the
> compilation,
> > shortly before code generation, rather than in make_hlds.m.
> >
> > > Anyone see problems with this approach or can think of another
> > > approach?
> >
> > There may be problems with "non-transparent" equivalent types:
> > those where a type parameter on the LHS does not occur on the RHS,
> > e.g.
> > :- type foo(T) == int.
> >
> > If you do the expansion before polymorphism, then the type_info for the
> > type parameter T won't get inserted in the arguments, but it needs to be.
>
> Are you sure that this is a problem.  Don't all occurences of foo(T) get
> transformed by equiv_type while still at the parse tree representation?  So
> the problem already exists.

Ah, you're right.  Currently we report

x.m:008: Sorry, not implemented: polymorphic equivalence type,
x.m:008:   with monomorphic definition, exported as abstract type.

So you don't need to worry about that case.

> > I think there will also be some issues with specialization and/or inlining
> > of unification procedures; the code there is probably assuming that the
> > argument types of the procedure will unify with the argument types in
> > the caller.
>
> But aren't the argument types in the caller of the expanded out equivalence
> type.  So before this change they didn't match but now they do.

Hmm, well, maybe you are right.  I guess you can go ahead and try it,
and see what (if anything) breaks.

I would advise testing it carefully, though;
in particular it's worth testing at -O5 as well as at -O2.

> I think I am unclear about when the expansion of equivalence types occurs.

Currently that happens immediately before make_hlds.

-- 
Fergus Henderson <fjh at 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.
--------------------------------------------------------------------------
mercury-developers mailing list
Post messages to:       mercury-developers at cs.mu.oz.au
Administrative Queries: owner-mercury-developers at cs.mu.oz.au
Subscriptions:          mercury-developers-request at cs.mu.oz.au
--------------------------------------------------------------------------