From: Christoph Lameter <clameter@sgi.com>
Subject: Page allocator: Optimize move_freepages() a bit
Move freepages has a strange loop without an increment and a bit of a complicated
control flow with two branches doing the same.
Make it more readable by specifying an increment. Also the skipping over pages is
improved if we use compound_order() to determine the order of allocated page.
Its likely helpful in particular if we have some huge pages around.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
---
mm/page_alloc.c | 23 +++++++++--------------
1 file changed, 9 insertions(+), 14 deletions(-)
Index: linux-2.6/mm/page_alloc.c
===================================================================
--- linux-2.6.orig/mm/page_alloc.c 2007-11-14 20:17:47.000000000 -0800
+++ linux-2.6/mm/page_alloc.c 2007-11-14 20:20:48.000000000 -0800
@@ -707,22 +707,17 @@ int move_freepages(struct zone *zone,
BUG_ON(page_zone(start_page) != page_zone(end_page));
#endif
- for (page = start_page; page <= end_page;) {
- if (!pfn_valid_within(page_to_pfn(page))) {
- page++;
- continue;
- }
+ for (page = start_page; page <= end_page; page += (1 << order)) {
- if (!PageBuddy(page)) {
- page++;
- continue;
- }
+ if (pfn_valid_within(page_to_pfn(page)) && PageBuddy(page)) {
- order = page_order(page);
- list_move(&page->lru,
- &zone->free_area[order].free_list[migratetype]);
- page += 1 << order;
- pages_moved += 1 << order;
+ order = page_order(page);
+ list_move(&page->lru,
+ &zone->free_area[order].free_list[migratetype]);
+ pages_moved += 1 << order;
+
+ } else
+ order = compound_order(page);
}
return pages_moved;