linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH] mm/zsmalloc: avoid duplicate assignment of prev_class
@ 2014-11-20 13:08 Mahendran Ganesh
  2014-11-21  3:22 ` Dan Streetman
  2014-11-21  3:43 ` Minchan Kim
  0 siblings, 2 replies; 5+ messages in thread
From: Mahendran Ganesh @ 2014-11-20 13:08 UTC (permalink / raw)
  To: minchan, ngupta, iamjoonsoo.kim; +Cc: linux-mm, linux-kernel, Mahendran Ganesh

In zs_create_pool(), prev_class is assigned (ZS_SIZE_CLASSES - 1)
times. And the prev_class only references to the previous alloc
size_class. So we do not need unnecessary assignement.

This patch modifies *prev_class* to *prev_alloc_class*. And the
*prev_alloc_class* will only be assigned when a new size_class
structure is allocated.

Signed-off-by: Mahendran Ganesh <opensource.ganesh@gmail.com>
---
 mm/zsmalloc.c |    9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/mm/zsmalloc.c b/mm/zsmalloc.c
index b3b57ef..ac2b396 100644
--- a/mm/zsmalloc.c
+++ b/mm/zsmalloc.c
@@ -970,7 +970,7 @@ struct zs_pool *zs_create_pool(gfp_t flags)
 		int size;
 		int pages_per_zspage;
 		struct size_class *class;
-		struct size_class *prev_class;
+		struct size_class *uninitialized_var(prev_alloc_class);
 
 		size = ZS_MIN_ALLOC_SIZE + i * ZS_SIZE_CLASS_DELTA;
 		if (size > ZS_MAX_ALLOC_SIZE)
@@ -987,9 +987,8 @@ struct zs_pool *zs_create_pool(gfp_t flags)
 		 * previous size_class if possible.
 		 */
 		if (i < ZS_SIZE_CLASSES - 1) {
-			prev_class = pool->size_class[i + 1];
-			if (can_merge(prev_class, size, pages_per_zspage)) {
-				pool->size_class[i] = prev_class;
+			if (can_merge(prev_alloc_class, size, pages_per_zspage)) {
+				pool->size_class[i] = prev_alloc_class;
 				continue;
 			}
 		}
@@ -1003,6 +1002,8 @@ struct zs_pool *zs_create_pool(gfp_t flags)
 		class->pages_per_zspage = pages_per_zspage;
 		spin_lock_init(&class->lock);
 		pool->size_class[i] = class;
+
+		prev_alloc_class = class;
 	}
 
 	pool->flags = flags;
-- 
1.7.9.5


^ permalink raw reply related	[flat|nested] 5+ messages in thread

* Re: [PATCH] mm/zsmalloc: avoid duplicate assignment of prev_class
  2014-11-20 13:08 [PATCH] mm/zsmalloc: avoid duplicate assignment of prev_class Mahendran Ganesh
@ 2014-11-21  3:22 ` Dan Streetman
  2014-11-21  5:22   ` Ganesh Mahendran
  2014-11-21  3:43 ` Minchan Kim
  1 sibling, 1 reply; 5+ messages in thread
From: Dan Streetman @ 2014-11-21  3:22 UTC (permalink / raw)
  To: Mahendran Ganesh
  Cc: Minchan Kim, Nitin Gupta, Joonsoo Kim, Linux-MM, linux-kernel

On Thu, Nov 20, 2014 at 8:08 AM, Mahendran Ganesh
<opensource.ganesh@gmail.com> wrote:
> In zs_create_pool(), prev_class is assigned (ZS_SIZE_CLASSES - 1)
> times. And the prev_class only references to the previous alloc
> size_class. So we do not need unnecessary assignement.
>
> This patch modifies *prev_class* to *prev_alloc_class*. And the
> *prev_alloc_class* will only be assigned when a new size_class
> structure is allocated.
>
> Signed-off-by: Mahendran Ganesh <opensource.ganesh@gmail.com>
> ---
>  mm/zsmalloc.c |    9 +++++----
>  1 file changed, 5 insertions(+), 4 deletions(-)
>
> diff --git a/mm/zsmalloc.c b/mm/zsmalloc.c
> index b3b57ef..ac2b396 100644
> --- a/mm/zsmalloc.c
> +++ b/mm/zsmalloc.c
> @@ -970,7 +970,7 @@ struct zs_pool *zs_create_pool(gfp_t flags)
>                 int size;
>                 int pages_per_zspage;
>                 struct size_class *class;
> -               struct size_class *prev_class;
> +               struct size_class *uninitialized_var(prev_alloc_class);

+               struct size_class *prev_class = NULL;

Use the fact it's unset below, so set it to NULL here

>
>                 size = ZS_MIN_ALLOC_SIZE + i * ZS_SIZE_CLASS_DELTA;
>                 if (size > ZS_MAX_ALLOC_SIZE)
> @@ -987,9 +987,8 @@ struct zs_pool *zs_create_pool(gfp_t flags)
>                  * previous size_class if possible.
>                  */
>                 if (i < ZS_SIZE_CLASSES - 1) {
> -                       prev_class = pool->size_class[i + 1];
> -                       if (can_merge(prev_class, size, pages_per_zspage)) {
> -                               pool->size_class[i] = prev_class;
> +                       if (can_merge(prev_alloc_class, size, pages_per_zspage)) {
> +                               pool->size_class[i] = prev_alloc_class;

simplify more, we can check if this is the first iteration by looking
at prev_class, e.g.:

                if (prev_class) {
                       if (can_merge(prev_class, size, pages_per_zspage)) {
                               pool->size_class[i] = prev_class;


>                                 continue;
>                         }
>                 }
> @@ -1003,6 +1002,8 @@ struct zs_pool *zs_create_pool(gfp_t flags)
>                 class->pages_per_zspage = pages_per_zspage;
>                 spin_lock_init(&class->lock);
>                 pool->size_class[i] = class;
> +
> +               prev_alloc_class = class;
>         }
>
>         pool->flags = flags;
> --
> 1.7.9.5
>
> --
> 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>

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] mm/zsmalloc: avoid duplicate assignment of prev_class
  2014-11-20 13:08 [PATCH] mm/zsmalloc: avoid duplicate assignment of prev_class Mahendran Ganesh
  2014-11-21  3:22 ` Dan Streetman
@ 2014-11-21  3:43 ` Minchan Kim
  2014-11-21  5:25   ` Ganesh Mahendran
  1 sibling, 1 reply; 5+ messages in thread
From: Minchan Kim @ 2014-11-21  3:43 UTC (permalink / raw)
  To: Mahendran Ganesh; +Cc: ngupta, iamjoonsoo.kim, linux-mm, linux-kernel

On Thu, Nov 20, 2014 at 09:08:33PM +0800, Mahendran Ganesh wrote:
> In zs_create_pool(), prev_class is assigned (ZS_SIZE_CLASSES - 1)
> times. And the prev_class only references to the previous alloc
> size_class. So we do not need unnecessary assignement.
> 
> This patch modifies *prev_class* to *prev_alloc_class*. And the
> *prev_alloc_class* will only be assigned when a new size_class
> structure is allocated.
> 
> Signed-off-by: Mahendran Ganesh <opensource.ganesh@gmail.com>
> ---
>  mm/zsmalloc.c |    9 +++++----
>  1 file changed, 5 insertions(+), 4 deletions(-)
> 
> diff --git a/mm/zsmalloc.c b/mm/zsmalloc.c
> index b3b57ef..ac2b396 100644
> --- a/mm/zsmalloc.c
> +++ b/mm/zsmalloc.c
> @@ -970,7 +970,7 @@ struct zs_pool *zs_create_pool(gfp_t flags)
>  		int size;
>  		int pages_per_zspage;
>  		struct size_class *class;
> -		struct size_class *prev_class;
> +		struct size_class *uninitialized_var(prev_alloc_class);

https://lkml.org/lkml/2012/10/27/71
In addition, I prefer prev_class.

Thanks.

-- 
Kind regards,
Minchan Kim

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] mm/zsmalloc: avoid duplicate assignment of prev_class
  2014-11-21  3:22 ` Dan Streetman
@ 2014-11-21  5:22   ` Ganesh Mahendran
  0 siblings, 0 replies; 5+ messages in thread
From: Ganesh Mahendran @ 2014-11-21  5:22 UTC (permalink / raw)
  To: Dan Streetman
  Cc: Minchan Kim, Nitin Gupta, Joonsoo Kim, Linux-MM, linux-kernel

Hello

2014-11-21 11:22 GMT+08:00 Dan Streetman <ddstreet@ieee.org>:
> On Thu, Nov 20, 2014 at 8:08 AM, Mahendran Ganesh
> <opensource.ganesh@gmail.com> wrote:
>> In zs_create_pool(), prev_class is assigned (ZS_SIZE_CLASSES - 1)
>> times. And the prev_class only references to the previous alloc
>> size_class. So we do not need unnecessary assignement.
>>
>> This patch modifies *prev_class* to *prev_alloc_class*. And the
>> *prev_alloc_class* will only be assigned when a new size_class
>> structure is allocated.
>>
>> Signed-off-by: Mahendran Ganesh <opensource.ganesh@gmail.com>
>> ---
>>  mm/zsmalloc.c |    9 +++++----
>>  1 file changed, 5 insertions(+), 4 deletions(-)
>>
>> diff --git a/mm/zsmalloc.c b/mm/zsmalloc.c
>> index b3b57ef..ac2b396 100644
>> --- a/mm/zsmalloc.c
>> +++ b/mm/zsmalloc.c
>> @@ -970,7 +970,7 @@ struct zs_pool *zs_create_pool(gfp_t flags)
>>                 int size;
>>                 int pages_per_zspage;
>>                 struct size_class *class;
>> -               struct size_class *prev_class;
>> +               struct size_class *uninitialized_var(prev_alloc_class);
>
> +               struct size_class *prev_class = NULL;
>
> Use the fact it's unset below, so set it to NULL here

Yes, You are right. I will change it in next resend.
Thanks for your review.

>
>>
>>                 size = ZS_MIN_ALLOC_SIZE + i * ZS_SIZE_CLASS_DELTA;
>>                 if (size > ZS_MAX_ALLOC_SIZE)
>> @@ -987,9 +987,8 @@ struct zs_pool *zs_create_pool(gfp_t flags)
>>                  * previous size_class if possible.
>>                  */
>>                 if (i < ZS_SIZE_CLASSES - 1) {
>> -                       prev_class = pool->size_class[i + 1];
>> -                       if (can_merge(prev_class, size, pages_per_zspage)) {
>> -                               pool->size_class[i] = prev_class;
>> +                       if (can_merge(prev_alloc_class, size, pages_per_zspage)) {
>> +                               pool->size_class[i] = prev_alloc_class;
>
> simplify more, we can check if this is the first iteration by looking
> at prev_class, e.g.:
>
>                 if (prev_class) {
>                        if (can_merge(prev_class, size, pages_per_zspage)) {
>                                pool->size_class[i] = prev_class;
>
>
>>                                 continue;
>>                         }
>>                 }
>> @@ -1003,6 +1002,8 @@ struct zs_pool *zs_create_pool(gfp_t flags)
>>                 class->pages_per_zspage = pages_per_zspage;
>>                 spin_lock_init(&class->lock);
>>                 pool->size_class[i] = class;
>> +
>> +               prev_alloc_class = class;
>>         }
>>
>>         pool->flags = flags;
>> --
>> 1.7.9.5
>>
>> --
>> 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>

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] mm/zsmalloc: avoid duplicate assignment of prev_class
  2014-11-21  3:43 ` Minchan Kim
@ 2014-11-21  5:25   ` Ganesh Mahendran
  0 siblings, 0 replies; 5+ messages in thread
From: Ganesh Mahendran @ 2014-11-21  5:25 UTC (permalink / raw)
  To: Minchan Kim; +Cc: Nitin Gupta, iamjoonsoo.kim, linux-mm, linux-kernel

Hello

2014-11-21 11:43 GMT+08:00 Minchan Kim <minchan@kernel.org>:
> On Thu, Nov 20, 2014 at 09:08:33PM +0800, Mahendran Ganesh wrote:
>> In zs_create_pool(), prev_class is assigned (ZS_SIZE_CLASSES - 1)
>> times. And the prev_class only references to the previous alloc
>> size_class. So we do not need unnecessary assignement.
>>
>> This patch modifies *prev_class* to *prev_alloc_class*. And the
>> *prev_alloc_class* will only be assigned when a new size_class
>> structure is allocated.
>>
>> Signed-off-by: Mahendran Ganesh <opensource.ganesh@gmail.com>
>> ---
>>  mm/zsmalloc.c |    9 +++++----
>>  1 file changed, 5 insertions(+), 4 deletions(-)
>>
>> diff --git a/mm/zsmalloc.c b/mm/zsmalloc.c
>> index b3b57ef..ac2b396 100644
>> --- a/mm/zsmalloc.c
>> +++ b/mm/zsmalloc.c
>> @@ -970,7 +970,7 @@ struct zs_pool *zs_create_pool(gfp_t flags)
>>               int size;
>>               int pages_per_zspage;
>>               struct size_class *class;
>> -             struct size_class *prev_class;
>> +             struct size_class *uninitialized_var(prev_alloc_class);
>
> https://lkml.org/lkml/2012/10/27/71
> In addition, I prefer prev_class.

Thanks for you review. I will resend the patch.

>
> Thanks.
>
> --
> Kind regards,
> Minchan Kim

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2014-11-21  5:25 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-11-20 13:08 [PATCH] mm/zsmalloc: avoid duplicate assignment of prev_class Mahendran Ganesh
2014-11-21  3:22 ` Dan Streetman
2014-11-21  5:22   ` Ganesh Mahendran
2014-11-21  3:43 ` Minchan Kim
2014-11-21  5:25   ` Ganesh Mahendran

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).