[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-mm
Subject:    Re: [Lhms-devel] Re: [RFC] buddy allocator without bitmap(2) [0/3]
From:       Hiroyuki KAMEZAWA <kamezawa.hiroyu () jp ! fujitsu ! com>
Date:       2004-08-31 22:44:32
Message-ID: 4134FF50.8000300 () jp ! fujitsu ! com
[Download RAW message or body]

Dave Hansen wrote:
> On Tue, 2004-08-31 at 03:36, Hiroyuki KAMEZAWA wrote:
> 
>>Disadvantage:
>>  - using one more PG_xxx flag.
>>  - If mem_map is not aligned, reserve one page as a victim for buddy allocater.
>>
>>How about this approach ?
> 
> 
> Granted, we have some free wiggle room in page->flags right now, but
> using another bit effectively shifts the entire benefit of your patch. 
> Instead of getting rid of the buddy bitmaps, you simply consume a
> page->flag instead.  While you don't have to allocate anything (because
> of the page->flags use), the number of bits consumed in the operation is
> still the same as before.  And the patch is getting more complex by the
> minute.

Hmm...I understand what you say. Consuming PG_xxx bit in buddy allocator is harmful
because no PG_xxx bit is used in current kernel's one.

What this patch implements is
"How to record shape of the mem_map needed by buddy allocator without using some
 structure which must be resized at memory resizing."

Because I had to record some information about shape of mem_map, I used PG_xxx bit.
1 bit is maybe minimum consumption.
If consumption of 1 bit in a page structure is too harmful,
I have to record information in some structure which needs to be resized
at memory resizing. When I do so, this patch itself is meaningless, I think.

I'll consider more, but...



> Something ate your patch:
> 
>    * Global page accounting.  One instance per CPU.  Only unsigned longs are
> @@ -290,6 +297,9 @@ extern unsigned long __read_page_state(u
>   #define SetPageCompound(page) set_bit(PG_compound, &(page)->flags)
>   #define ClearPageCompound(page)       clear_bit(PG_compound, &(page)->flags)
> 
> +#define PageBuddyend(page)      test_bit(PG_buddyend, &(page)->flags)
> +#define SetPageBuddyend(page)   set_bit(PG_buddyend, &(page)->flags)
> +
>   #ifdef CONFIG_SWAP
>   #define PageSwapCache(page)   test_bit(PG_swapcache, &(page)->flags)
>   #define SetPageSwapCache(page)        set_bit(PG_swapcache, &(page)->flags)
> 
> 
> -- Dave


-- 
--the clue is these footmarks leading to the door.--
KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>

--
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:"aart@kvack.org"> aart@kvack.org </a>
[prev in list] [next in list] [prev in thread] [next in thread]