[PATCH] memory page_alloc zonelist caching speedup aligncache

[PATCH] memory page_alloc zonelist caching speedup aligncache

[Paul Jackson, Paul Jackson]

Avoid frequent writes to the zonelist zones[] array, which are
read-only after initial setup, by putting the zonelist_cache on
a separate cacheline.

Signed-off-by: Paul Jackson <pj@sgi.com>

---

 include/linux/mmzone.h |    3 ++-
 1 files changed, 2 insertions(+), 1 deletion(-)

--- 2.6.19-rc2-mm1.orig/include/linux/mmzone.h	2006-10-17 17:19:22.000000000 -0700
+++ 2.6.19-rc2-mm1/include/linux/mmzone.h	2006-10-17 17:31:31.000000000 -0700
@@ -396,7 +396,8 @@ struct zonelist {
 	struct zonelist_cache *zlcache_ptr;		     // NULL or &zlcache
 	struct zone *zones[MAX_ZONES_PER_ZONELIST + 1];      // NULL delimited
 #ifdef CONFIG_NUMA
-	struct zonelist_cache zlcache;			     // optional ...
+	/* Keep written zonelist_cache off read-only zones[] cache lines */
+	struct zonelist_cache zlcache ____cacheline_aligned; // optional ...
 #endif
 };
 

-- 
                  I won't rest till it's the best ...
                  Programmer, Linux Scalability
                  Paul Jackson <pj@sgi.com> 1.925.600.0401

--
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] memory page_alloc zonelist caching speedup aligncache

[Rohit Seth]

On Wed, 2006-10-18 at 01:14 -0700, Paul Jackson wrote:
> From: Paul Jackson <pj@sgi.com>
> 
> Avoid frequent writes to the zonelist zones[] array, which are
> read-only after initial setup, by putting the zonelist_cache on
> a separate cacheline.
> 
> Signed-off-by: Paul Jackson <pj@sgi.com>
> 
> ---
> 
>  include/linux/mmzone.h |    3 ++-
>  1 files changed, 2 insertions(+), 1 deletion(-)
> 
> --- 2.6.19-rc2-mm1.orig/include/linux/mmzone.h	2006-10-17 17:19:22.000000000 -0700
> +++ 2.6.19-rc2-mm1/include/linux/mmzone.h	2006-10-17 17:31:31.000000000 -0700
> @@ -396,7 +396,8 @@ struct zonelist {
>  	struct zonelist_cache *zlcache_ptr;		     // NULL or &zlcache
>  	struct zone *zones[MAX_ZONES_PER_ZONELIST + 1];      // NULL delimited
>  #ifdef CONFIG_NUMA
> -	struct zonelist_cache zlcache;			     // optional ...
> +	/* Keep written zonelist_cache off read-only zones[] cache lines */
> +	struct zonelist_cache zlcache ____cacheline_aligned; // optional ...
>  #endif
>  };
>  
> 

Wouldn't it be better to have the read mostly field z_to_n defined first
in zonelist_cache (and define the BITMAP at the end).

-rohit

--
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>