This is the mail archive of the
mailing list for the GCC project.
Re: RFA: patch to solve PR37859
Vladimir Makarov wrote:
Adam Nemet wrote:
Vlad,Here is the modified version of the patch. I also decided to use a
new attribute for ira_loop_tree_node because calculation of it for
each store is a bit more expensive.
Thanks for patch. I think I understand the difference. What you're
that the reason the change on r719 does not reach r696 because the
r719 flows directly into an adjacent loop instead of the parent loop.
I will bootstrap the patch on mips64 overnight and report back.
There are few comments below FWIW.
2008-11-14 Vladimir Makarov <firstname.lastname@example.org>
* ira-int.h (struct ira_loop_tree_node): New member
* ira-color.c (print_loop_title): Print loop bbs.
* ira-emit.c (entered_from_non_parent_p,
setup_entered_from_non_parent_p): New functions.
(not_modified_p): Rename to store_can_be_removed_p. Check there
is no side entries.
(generate_edge_moves): Use store_can_be_removed_p instead of
(ira_emit): Call setup_entered_from_non_parent_p.
* ira-build.c (copy_info_to_removed_store_destinations):
Accumulate CALL_FREQ, CALL_CROSSED_NUM, and
(ira_flattening): Don't CHECK MEM_OPTIMIZED_DEST[_P], always
update all accumulated attributes.