This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [lno] Use scev in ivopts
Hello,
> I still don't understand why after having determined the scalar
> evolution with analyze_scalar_evolution, you don't use
> instantiate_parameters . Furthermore I get strange ICEs if I try to
> insert calls to instantiate_parameters.
>
> On Thu, Apr 15, 2004 at 12:36:48AM +0200, Zdenek Dvorak wrote:
>
> > --- 1045,1064 ----
> > && TREE_CODE (type) != POINTER_TYPE)
> > return NULL_TREE;
> >
> > ! ev = analyze_scalar_evolution (loop, name);
>
> here, I would have used:
>
> ev = instantiate_parameters (loop, analyze_scalar_evolution (loop, name));
why? What I want is to get base + step * i, where base is expressed
purely in loop invariants (which is what analyze_scalar_evolution
returns now by default). Once the symbolic parameters work again, I
will need to add a check here that no symbolic parameters occur in the
base, but that's it -- I don't need to instantiate any parameters.
Zdenek