This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH][C++/Java] Get rid of the walk_subtrees langhook
- From: Andrew Haley <aph-gcc at littlepinkcloud dot COM>
- To: Richard Guenther <rguenther at suse dot de>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Thu, 26 Jul 2007 17:17:08 +0100
- Subject: Re: [PATCH][C++/Java] Get rid of the walk_subtrees langhook
- References: <Pine.LNX.firstname.lastname@example.org>
Richard Guenther writes:
> This gets rid of the walk_subtrees langhook which is called at every
> invocation of walk_trees. The only real user is the C++ frontend which
> is too lazy to use its own walk_tree. Instead of duplicating walk_tree
> there I made the langhook an extra callback parameter for walk_tree
> and replaced all occurances in the C++ frontend by a wrapper called
> cp_walk_tree that passes the current langhook implementation. All
> cases covered in the langhook look frontend specific and should not
> affect middle-end walkers.
> Apart from nice to LTO this should bring a noticable speedup for
> targets with slow indirect calls.
> Full bootstrap and regtest on x86_64-unknown-linux-gnu still running
> for C++ and Java, though build and C++ testing finished successfully
> Ok for mainline? (This needs a GWP at best)
> Note I did not go through callers to walk_tree in the C++ frontend
> to identify which ones really need the callback but just changed them
> all over.
I guess that's OK, yes. Now that inlining is done on generic trees
rather than java FE trees there's no need for java_tree_inlining_walk_subtrees.