2022-04-16 09:23:59

by Wonhyuk Yang

[permalink] [raw]
Subject: [PATCH v2] mm/slub: Remove repeated action in calculate_order()

To calculate order, calc_slab_order() is called repeatly changing the
fract_leftover. Thus, the branch which is not dependent on
fract_leftover is executed repeatly. So make it run only once.

Plus, when min_object reached to 0, we set fract_leftover to 1. In
this case, we can calculate order by max(slub_min_order,
get_order(size)) instead of calling calc_slab_order().

No functional impact expected.

Signed-off-by: Wonhyuk Yang <[email protected]>
---
V1 -> V2: Fix typo miss in a commit message

mm/slub.c | 18 +++++++-----------
1 file changed, 7 insertions(+), 11 deletions(-)

diff --git a/mm/slub.c b/mm/slub.c
index ed5c2c03a47a..e7a394d7b75a 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -3795,9 +3795,6 @@ static inline unsigned int calc_slab_order(unsigned int size,
unsigned int min_order = slub_min_order;
unsigned int order;

- if (order_objects(min_order, size) > MAX_OBJS_PER_PAGE)
- return get_order(size * MAX_OBJS_PER_PAGE) - 1;
-
for (order = max(min_order, (unsigned int)get_order(min_objects * size));
order <= max_order; order++) {

@@ -3820,6 +3817,11 @@ static inline int calculate_order(unsigned int size)
unsigned int max_objects;
unsigned int nr_cpus;

+ if (unlikely(order_objects(slub_min_order, size) > MAX_OBJS_PER_PAGE)) {
+ order = get_order(size * MAX_OBJS_PER_PAGE) - 1;
+ goto out;
+ }
+
/*
* Attempt to find best configuration for a slab. This
* works by first attempting to generate a layout with
@@ -3865,14 +3867,8 @@ static inline int calculate_order(unsigned int size)
* We were unable to place multiple objects in a slab. Now
* lets see if we can place a single object there.
*/
- order = calc_slab_order(size, 1, slub_max_order, 1);
- if (order <= slub_max_order)
- return order;
-
- /*
- * Doh this slab cannot be placed using slub_max_order.
- */
- order = calc_slab_order(size, 1, MAX_ORDER, 1);
+ order = max_t(unsigned int, slub_min_order, (unsigned int)get_order(size));
+out:
if (order < MAX_ORDER)
return order;
return -ENOSYS;
--
2.30.2


2022-04-18 04:07:05

by Hyeonggon Yoo

[permalink] [raw]
Subject: Re: [PATCH v2] mm/slub: Remove repeated action in calculate_order()

On Sat, Apr 16, 2022 at 04:40:59PM +0900, Wonhyuk Yang wrote:
> To calculate order, calc_slab_order() is called repeatly changing the
> fract_leftover. Thus, the branch which is not dependent on
> fract_leftover is executed repeatly. So make it run only once.
>
> Plus, when min_object reached to 0, we set fract_leftover to 1. In

Maybe you mean when min_object reached 1.

> this case, we can calculate order by max(slub_min_order,
> get_order(size)) instead of calling calc_slab_order().
>
> No functional impact expected.
> Signed-off-by: Wonhyuk Yang <[email protected]>
> ---
> V1 -> V2: Fix typo miss in a commit message
>
> mm/slub.c | 18 +++++++-----------
> 1 file changed, 7 insertions(+), 11 deletions(-)
>
> diff --git a/mm/slub.c b/mm/slub.c
> index ed5c2c03a47a..e7a394d7b75a 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -3795,9 +3795,6 @@ static inline unsigned int calc_slab_order(unsigned int size,
> unsigned int min_order = slub_min_order;
> unsigned int order;
>
> - if (order_objects(min_order, size) > MAX_OBJS_PER_PAGE)
> - return get_order(size * MAX_OBJS_PER_PAGE) - 1;
> -
> for (order = max(min_order, (unsigned int)get_order(min_objects * size));
> order <= max_order; order++) {
>
> @@ -3820,6 +3817,11 @@ static inline int calculate_order(unsigned int size)
> unsigned int max_objects;
> unsigned int nr_cpus;
>
> + if (unlikely(order_objects(slub_min_order, size) > MAX_OBJS_PER_PAGE)) {
> + order = get_order(size * MAX_OBJS_PER_PAGE) - 1;
> + goto out;
> + }
> +
> /*
> * Attempt to find best configuration for a slab. This
> * works by first attempting to generate a layout with
> @@ -3865,14 +3867,8 @@ static inline int calculate_order(unsigned int size)
> * We were unable to place multiple objects in a slab. Now
> * lets see if we can place a single object there.
> */
> - order = calc_slab_order(size, 1, slub_max_order, 1);
> - if (order <= slub_max_order)
> - return order;
> -
> - /*
> - * Doh this slab cannot be placed using slub_max_order.
> - */
> - order = calc_slab_order(size, 1, MAX_ORDER, 1);
> + order = max_t(unsigned int, slub_min_order, (unsigned int)get_order(size));
> +out:

You don't need to cast value of get_order(size). max_t() does cast both operands.

> if (order < MAX_ORDER)
> return order;
> return -ENOSYS;

For the correctness of the patch, I don't see any problem about the
code.

But to be honest I'm a bit skeptical about saving some cycles in
calculating slab order. It's done only when creating caches (usually in boot
process).

> --
> 2.30.2
>
>

--
Thanks,
Hyeonggon

2022-04-18 05:01:06

by Wonhyuk Yang

[permalink] [raw]
Subject: Re: [PATCH v2] mm/slub: Remove repeated action in calculate_order()

On Sun, Apr 17, 2022 at 10:43 AM Hyeonggon Yoo <[email protected]> wrote:
>
> On Sat, Apr 16, 2022 at 04:40:59PM +0900, Wonhyuk Yang wrote:
> > To calculate order, calc_slab_order() is called repeatly changing the
> > fract_leftover. Thus, the branch which is not dependent on
> > fract_leftover is executed repeatly. So make it run only once.
> >
> > Plus, when min_object reached to 0, we set fract_leftover to 1. In
>
> Maybe you mean when min_object reached 1.

Yes, That comment need to be updated...

>
> > this case, we can calculate order by max(slub_min_order,
> > get_order(size)) instead of calling calc_slab_order().
> >
> > No functional impact expected.
> > Signed-off-by: Wonhyuk Yang <[email protected]>
> > ---
> > V1 -> V2: Fix typo miss in a commit message
> >
> > mm/slub.c | 18 +++++++-----------
> > 1 file changed, 7 insertions(+), 11 deletions(-)
> >
> > diff --git a/mm/slub.c b/mm/slub.c
> > index ed5c2c03a47a..e7a394d7b75a 100644
> > --- a/mm/slub.c
> > +++ b/mm/slub.c
> > @@ -3795,9 +3795,6 @@ static inline unsigned int calc_slab_order(unsigned int size,
> > unsigned int min_order = slub_min_order;
> > unsigned int order;
> >
> > - if (order_objects(min_order, size) > MAX_OBJS_PER_PAGE)
> > - return get_order(size * MAX_OBJS_PER_PAGE) - 1;
> > -
> > for (order = max(min_order, (unsigned int)get_order(min_objects * size));
> > order <= max_order; order++) {
> >
> > @@ -3820,6 +3817,11 @@ static inline int calculate_order(unsigned int size)
> > unsigned int max_objects;
> > unsigned int nr_cpus;
> >
> > + if (unlikely(order_objects(slub_min_order, size) > MAX_OBJS_PER_PAGE)) {
> > + order = get_order(size * MAX_OBJS_PER_PAGE) - 1;
> > + goto out;
> > + }
> > +
> > /*
> > * Attempt to find best configuration for a slab. This
> > * works by first attempting to generate a layout with
> > @@ -3865,14 +3867,8 @@ static inline int calculate_order(unsigned int size)
> > * We were unable to place multiple objects in a slab. Now
> > * lets see if we can place a single object there.
> > */
> > - order = calc_slab_order(size, 1, slub_max_order, 1);
> > - if (order <= slub_max_order)
> > - return order;
> > -
> > - /*
> > - * Doh this slab cannot be placed using slub_max_order.
> > - */
> > - order = calc_slab_order(size, 1, MAX_ORDER, 1);
> > + order = max_t(unsigned int, slub_min_order, (unsigned int)get_order(size));
> > +out:
>
> You don't need to cast value of get_order(size). max_t() does cast both operands.

That's a good point, I will delete it.

>
> > if (order < MAX_ORDER)
> > return order;
> > return -ENOSYS;
>
> For the correctness of the patch, I don't see any problem about the
> code.
>
> But to be honest I'm a bit skeptical about saving some cycles in
> calculating slab order. It's done only when creating caches (usually in boot
> process).
>

2022-04-18 07:20:40

by Hyeonggon Yoo

[permalink] [raw]
Subject: Re: [PATCH v2] mm/slub: Remove repeated action in calculate_order()

On Sun, Apr 17, 2022 at 10:43:20AM +0900, Hyeonggon Yoo wrote:
> On Sat, Apr 16, 2022 at 04:40:59PM +0900, Wonhyuk Yang wrote:
> > To calculate order, calc_slab_order() is called repeatly changing the
> > fract_leftover. Thus, the branch which is not dependent on
> > fract_leftover is executed repeatly. So make it run only once.
> >
> > Plus, when min_object reached to 0, we set fract_leftover to 1. In
>
> Maybe you mean when min_object reached 1.
>
> > this case, we can calculate order by max(slub_min_order,
> > get_order(size)) instead of calling calc_slab_order().
> >
> > No functional impact expected.
> > Signed-off-by: Wonhyuk Yang <[email protected]>
> > ---
> > V1 -> V2: Fix typo miss in a commit message
> >
> > mm/slub.c | 18 +++++++-----------
> > 1 file changed, 7 insertions(+), 11 deletions(-)
> >
> > diff --git a/mm/slub.c b/mm/slub.c
> > index ed5c2c03a47a..e7a394d7b75a 100644
> > --- a/mm/slub.c
> > +++ b/mm/slub.c
> > @@ -3795,9 +3795,6 @@ static inline unsigned int calc_slab_order(unsigned int size,
> > unsigned int min_order = slub_min_order;
> > unsigned int order;
> >
> > - if (order_objects(min_order, size) > MAX_OBJS_PER_PAGE)
> > - return get_order(size * MAX_OBJS_PER_PAGE) - 1;
> > -
> > for (order = max(min_order, (unsigned int)get_order(min_objects * size));
> > order <= max_order; order++) {
> >
> > @@ -3820,6 +3817,11 @@ static inline int calculate_order(unsigned int size)
> > unsigned int max_objects;
> > unsigned int nr_cpus;
> >
> > + if (unlikely(order_objects(slub_min_order, size) > MAX_OBJS_PER_PAGE)) {
> > + order = get_order(size * MAX_OBJS_PER_PAGE) - 1;
> > + goto out;
> > + }
> > +
> > /*
> > * Attempt to find best configuration for a slab. This
> > * works by first attempting to generate a layout with
> > @@ -3865,14 +3867,8 @@ static inline int calculate_order(unsigned int size)
> > * We were unable to place multiple objects in a slab. Now
> > * lets see if we can place a single object there.
> > */
> > - order = calc_slab_order(size, 1, slub_max_order, 1);
> > - if (order <= slub_max_order)
> > - return order;
> > -
> > - /*
> > - * Doh this slab cannot be placed using slub_max_order.
> > - */
> > - order = calc_slab_order(size, 1, MAX_ORDER, 1);
> > + order = max_t(unsigned int, slub_min_order, (unsigned int)get_order(size));
> > +out:
>
> You don't need to cast value of get_order(size). max_t() does cast both operands.
>
> > if (order < MAX_ORDER)
> > return order;
> > return -ENOSYS;
>
> For the correctness of the patch, I don't see any problem about the
> code.
>
> But to be honest I'm a bit skeptical about saving some cycles in
> calculating slab order. It's done only when creating caches (usually in boot
> process).

But yeah, maybe it's worth for better maintenance of code.

So after considering my comments, feel free to add:
Reviewed-by: Hyeonggon Yoo <[email protected]>

Thanks!

> > --
> > 2.30.2
> >
> >
>
> --
> Thanks,
> Hyeonggon

--
Thanks,
Hyeonggon

2022-04-18 12:22:20

by Wonhyuk Yang

[permalink] [raw]
Subject: Re: [PATCH v2] mm/slub: Remove repeated action in calculate_order()

On Sun, Apr 17, 2022 at 11:07 AM Hyeonggon Yoo <[email protected]> wrote:
>
> On Sun, Apr 17, 2022 at 10:43:20AM +0900, Hyeonggon Yoo wrote:
> > On Sat, Apr 16, 2022 at 04:40:59PM +0900, Wonhyuk Yang wrote:
> > > To calculate order, calc_slab_order() is called repeatly changing the
> > > fract_leftover. Thus, the branch which is not dependent on
> > > fract_leftover is executed repeatly. So make it run only once.
> > >
> > > Plus, when min_object reached to 0, we set fract_leftover to 1. In
> >
> > Maybe you mean when min_object reached 1.
> >
> > > this case, we can calculate order by max(slub_min_order,
> > > get_order(size)) instead of calling calc_slab_order().
> > >
> > > No functional impact expected.
> > > Signed-off-by: Wonhyuk Yang <[email protected]>
> > > ---
> > > V1 -> V2: Fix typo miss in a commit message
> > >
> > > mm/slub.c | 18 +++++++-----------
> > > 1 file changed, 7 insertions(+), 11 deletions(-)
> > >
> > > diff --git a/mm/slub.c b/mm/slub.c
> > > index ed5c2c03a47a..e7a394d7b75a 100644
> > > --- a/mm/slub.c
> > > +++ b/mm/slub.c
> > > @@ -3795,9 +3795,6 @@ static inline unsigned int calc_slab_order(unsigned int size,
> > > unsigned int min_order = slub_min_order;
> > > unsigned int order;
> > >
> > > - if (order_objects(min_order, size) > MAX_OBJS_PER_PAGE)
> > > - return get_order(size * MAX_OBJS_PER_PAGE) - 1;
> > > -
> > > for (order = max(min_order, (unsigned int)get_order(min_objects * size));
> > > order <= max_order; order++) {
> > >
> > > @@ -3820,6 +3817,11 @@ static inline int calculate_order(unsigned int size)
> > > unsigned int max_objects;
> > > unsigned int nr_cpus;
> > >
> > > + if (unlikely(order_objects(slub_min_order, size) > MAX_OBJS_PER_PAGE)) {
> > > + order = get_order(size * MAX_OBJS_PER_PAGE) - 1;
> > > + goto out;
> > > + }
> > > +
> > > /*
> > > * Attempt to find best configuration for a slab. This
> > > * works by first attempting to generate a layout with
> > > @@ -3865,14 +3867,8 @@ static inline int calculate_order(unsigned int size)
> > > * We were unable to place multiple objects in a slab. Now
> > > * lets see if we can place a single object there.
> > > */
> > > - order = calc_slab_order(size, 1, slub_max_order, 1);
> > > - if (order <= slub_max_order)
> > > - return order;
> > > -
> > > - /*
> > > - * Doh this slab cannot be placed using slub_max_order.
> > > - */
> > > - order = calc_slab_order(size, 1, MAX_ORDER, 1);
> > > + order = max_t(unsigned int, slub_min_order, (unsigned int)get_order(size));
> > > +out:
> >
> > You don't need to cast value of get_order(size). max_t() does cast both operands.
> >
> > > if (order < MAX_ORDER)
> > > return order;
> > > return -ENOSYS;
> >
> > For the correctness of the patch, I don't see any problem about the
> > code.
> >
> > But to be honest I'm a bit skeptical about saving some cycles in
> > calculating slab order. It's done only when creating caches (usually in boot
> > process).
>
> But yeah, maybe it's worth for better maintenance of code.
>
> So after considering my comments, feel free to add:
> Reviewed-by: Hyeonggon Yoo <[email protected]>
>
> Thanks!
>

Thanks for taking the time to review Hyeonggon!