[PATCH] mm : kill combined_idx

[PATCH] mm : kill combined_idx

[Huang Shijie]

In more then half of all the cases, `page' is head of the buddy pair
{page, buddy} in __free_one_page. That is because the allocation logic
always picks the head of a chunk, and puts the rest back to the buddy system.

So calculating the combined page is not needed but waste some cycles in
more then half of all the cases.Just do the calculation when `page' is
bigger then the `buddy'.

Signed-off-by: Huang Shijie <shijie8@gmail.com>
---
 mm/page_alloc.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 4e86965..42351bf 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -464,7 +464,6 @@ static inline void __free_one_page(struct page *page,
 	VM_BUG_ON(bad_range(zone, page));
 
 	while (order < MAX_ORDER-1) {
-		unsigned long combined_idx;
 		struct page *buddy;
 
 		buddy = __page_find_buddy(page, page_idx, order);
@@ -475,9 +474,10 @@ static inline void __free_one_page(struct page *page,
 		list_del(&buddy->lru);
 		zone->free_area[order].nr_free--;
 		rmv_page_order(buddy);
-		combined_idx = __find_combined_index(page_idx, order);
-		page = page + (combined_idx - page_idx);
-		page_idx = combined_idx;
+		if (page > buddy) { /* keep `page' the head of the buddy pair */
+			page = buddy;
+			page_idx = __find_combined_index(page_idx, order);
+		}
 		order++;
 	}
 	set_page_order(page, order);
-- 
1.6.0.6

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] mm : kill combined_idx

[KAMEZAWA Hiroyuki]

On Mon, 21 Dec 2009 11:32:27 +0800
Huang Shijie <shijie8@gmail.com> wrote:

> In more then half of all the cases, `page' is head of the buddy pair
> {page, buddy} in __free_one_page. That is because the allocation logic
> always picks the head of a chunk, and puts the rest back to the buddy system.
> 
> So calculating the combined page is not needed but waste some cycles in
> more then half of all the cases.Just do the calculation when `page' is
> bigger then the `buddy'.
> 
> Signed-off-by: Huang Shijie <shijie8@gmail.com>

Hmm...As far as I remember, this code design was for avoiding "if".
Is this compare+jump is better than add+xor ?

Thanks,
-Kame

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] mm : kill combined_idx

[Mel Gorman]

On Mon, Dec 21, 2009 at 02:31:39PM +0900, KAMEZAWA Hiroyuki wrote:
> On Mon, 21 Dec 2009 11:32:27 +0800
> Huang Shijie <shijie8@gmail.com> wrote:
> 
> > In more then half of all the cases, `page' is head of the buddy pair
> > {page, buddy} in __free_one_page. That is because the allocation logic
> > always picks the head of a chunk, and puts the rest back to the buddy system.
> > 
> > So calculating the combined page is not needed but waste some cycles in
> > more then half of all the cases.Just do the calculation when `page' is
> > bigger then the `buddy'.
> > 
> > Signed-off-by: Huang Shijie <shijie8@gmail.com>
> 
> Hmm...As far as I remember, this code design was for avoiding "if".
> Is this compare+jump is better than add+xor ?
> 

Agreed. It's not clear that a compare+jump is cheaper than the add+xor.
How often it's the case that the page is the higher or lower half of the
buddy would depend heavily on the allocation/free pattern making it
hard, if not possible, to predict which is the more common case.

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] mm : kill combined_idx

[Huang Shijie]

>> Hmm...As far as I remember, this code design was for avoiding "if".
>> Is this compare+jump is better than add+xor ?
>>
>>     
>
> Agreed. It's not clear that a compare+jump is cheaper than the add+xor.
> How often it's the case that the page is the higher or lower half of the
> buddy would depend heavily on the allocation/free pattern making it
> hard, if not possible, to predict which is the more common case.
>
>   
ok. thanks a lot.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>