This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Fix PR17687, sincos causing extra addressable vars
On Fri, 8 Dec 2006, Andrew Pinski wrote:
> On Thu, 2006-12-07 at 13:50 +0100, Richard Guenther wrote:
>
> This part:
> > + tree fn;
> > +
> > + /* We treat sincos specially here because we want to avoid
> > + making the second and third argument addressable. We do so
> > + by creating a new compound expression with a temporary
> > + complex value to store the result of a cexpi call
> > + and to initialize the sincos result arguments. */
>
> Seems like a hack, a very very bad one. I would rather have the
> gimplifier just do that gimplify and not transform stuff like this.
> This really belongs in builtins.c in something like fold_sincos.
> In builtins.c I would use a SAVE_EXPR and that solves the problem of
> having to creating a temporary variable.
I don't agree. I had the fold variant before and people didn't like it.
Also folding will result in non-gimple which means I could as well
stick it into convert.c. Also I don't see how SAVE_EXPR will help,
can you elaborate? (ok, I guess we can share SAVE_EXPR trees?)
Richard.
--
Richard Guenther <rguenther@suse.de>
Novell / SUSE Labs