This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patch] Allow merging blocks with phi nodes
On Mon, 2005-05-16 at 16:16, Zdenek Dvorak wrote:
> Hello,
>
> > > New version of fold_stmt (fold_stmt_inplace) is necessary, since
> > > fold_stmt may replace the whole statement, but we cannot get the pointer
> > > to statement from immediate uses, only the statement itself.
> >
> > Isn't that what bsi_for_stmt () is for?
>
> bsi_for_stmt is slow at the moment (and fold_stmt does not do much more
> than fold_stmt_inplace does, anyway).
>
Wasn't Kazu working on a patch to fold the annotations into the
statement_list node? That was going to provide the O(1) bsi_for_stmt
because the stmt annotation pointer was going to be replaced with the
statement_list pointer.
Or did that not completely get off the ground? :-)
Andrew