This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [tuples] [patch] refactoring of tree_expr_nonzero_warnv_p
On Tue, Mar 4, 2008 at 1:47 PM, Diego Novillo <dnovillo@google.com> wrote:
> On 3/4/08 5:38 AM, Richard Guenther wrote:
>
> > Please use context diffs when posting patches, and do them inline
> > rather than with attachments. Thanks.
>
> Ironically, I'd rather have attached unified diffs. I guess it's a
> matter of taste.
Well, unified diffs usually "optimize" for size and so get removes/adds
of different functions cluttered all over the place which makes the patch
hard to follow. So I prefer context diffs.
> > I find the _expr bits in the function names redundant,
> > tree_unary_nonzero_warnv_p is shorter and as explicit. It also
> > matches
> > fold_unary and friends.
>
> Yeah, that's probably a good idea.
>
>
> > IMHO all this mixing of GIMPLE concepts inside fold-const.c is ugly,
> > can't we just cp fold-const.c gimple-fold-const.c and only fix the duplicate?
> > (Yeah, I hope we can trim down both sides and finally get the FEs only
> > fold what they really need).
>
> Yes, but the gimple counterparts will need the basic functionality in
> fold-const.c. There will be a gimple-fold.c soon. We are implementing
> GIMPLE folders in terms of public functions in fold-const.c, which is
> shared by the FEs and the ME.
Ok.
> So, in this case I propose moving the tree_*_nonzero_warnv_p() functions
> to fold-const.c and gimple_nonzero_warnv_p() in gimple-fold.c.
> Similarly, for other functions that we have in tree-ssa-ccp.c and
> tree-ssa.c. Thoughts?
Sounds like a good plan. Maybe also separate constant foldings
(that would be also used by FEs directly) to a const-fold.c. But that's
of course unrelated to your work here.
Richard.