2017-08-03 06:38:28

by Wei Yang

Subject: [PATCH] mm/vmalloc: reduce half comparison during pcpu_get_vm_areas()

In pcpu_get_vm_areas(), it checks each range is not overlapped. To make
sure it is, only (N^2)/2 comparison is necessary, while current code does
N^2 times. By starting from the next range, it achieves the goal and the
continue could be removed.

At the mean time, other two work in this patch:
* the overlap check of two ranges could be done with one clause
* one typo in comment is fixed.

Signed-off-by: Wei Yang <[email protected]>
---
mm/vmalloc.c | 10 +++-------
1 file changed, 3 insertions(+), 7 deletions(-)

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index 8087451cb332..f33c8350fd83 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -2457,7 +2457,7 @@ static unsigned long pvm_determine_end(struct vmap_area **pnext,
* matching slot. While scanning, if any of the areas overlaps with
* existing vmap_area, the base address is pulled down to fit the
* area. Scanning is repeated till all the areas fit and then all
- * necessary data structres are inserted and the result is returned.
+ * necessary data structures are inserted and the result is returned.
*/
struct vm_struct **pcpu_get_vm_areas(const unsigned long *offsets,
const size_t *sizes, int nr_vms,
@@ -2485,15 +2485,11 @@ struct vm_struct **pcpu_get_vm_areas(const unsigned long *offsets,
if (start > offsets[last_area])
last_area = area;

- for (area2 = 0; area2 < nr_vms; area2++) {
+ for (area2 = area + 1; area2 < nr_vms; area2++) {
unsigned long start2 = offsets[area2];
unsigned long end2 = start2 + sizes[area2];

- if (area2 == area)
- continue;
-
- BUG_ON(start2 >= start && start2 < end);
- BUG_ON(end2 <= end && end2 > start);
+ BUG_ON(start2 < end && start < end2);
}
}
last_end = offsets[last_area] + sizes[last_area];
--
2.11.0

2017-08-07 11:40:47

by Michal Hocko

Subject: Re: [PATCH] mm/vmalloc: reduce half comparison during pcpu_get_vm_areas()

[CC Tejun]

On Thu 03-08-17 14:38:22, Wei Yang wrote:
> In pcpu_get_vm_areas(), it checks each range is not overlapped. To make
> sure it is, only (N^2)/2 comparison is necessary, while current code does
> N^2 times. By starting from the next range, it achieves the goal and the
> continue could be removed.
>
> At the mean time, other two work in this patch:
> * the overlap check of two ranges could be done with one clause
> * one typo in comment is fixed.
>
> Signed-off-by: Wei Yang <[email protected]>
> ---
> mm/vmalloc.c | 10 +++-------
> 1 file changed, 3 insertions(+), 7 deletions(-)
>
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index 8087451cb332..f33c8350fd83 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -2457,7 +2457,7 @@ static unsigned long pvm_determine_end(struct vmap_area **pnext,
> * matching slot. While scanning, if any of the areas overlaps with
> * existing vmap_area, the base address is pulled down to fit the
> * area. Scanning is repeated till all the areas fit and then all
> - * necessary data structres are inserted and the result is returned.
> + * necessary data structures are inserted and the result is returned.
> */
> struct vm_struct **pcpu_get_vm_areas(const unsigned long *offsets,
> const size_t *sizes, int nr_vms,
> @@ -2485,15 +2485,11 @@ struct vm_struct **pcpu_get_vm_areas(const unsigned long *offsets,
> if (start > offsets[last_area])
> last_area = area;
>
> - for (area2 = 0; area2 < nr_vms; area2++) {
> + for (area2 = area + 1; area2 < nr_vms; area2++) {
> unsigned long start2 = offsets[area2];
> unsigned long end2 = start2 + sizes[area2];
>
> - if (area2 == area)
> - continue;
> -
> - BUG_ON(start2 >= start && start2 < end);
> - BUG_ON(end2 <= end && end2 > start);
> + BUG_ON(start2 < end && start < end2);
> }
> }
> last_end = offsets[last_area] + sizes[last_area];
> --
> 2.11.0
>

--
Michal Hocko
SUSE Labs

2017-08-07 15:11:09

by Tejun Heo

Subject: Re: [PATCH] mm/vmalloc: reduce half comparison during pcpu_get_vm_areas()

On Thu, Aug 03, 2017 at 02:38:22PM +0800, Wei Yang wrote:
> In pcpu_get_vm_areas(), it checks each range is not overlapped. To make
> sure it is, only (N^2)/2 comparison is necessary, while current code does
> N^2 times. By starting from the next range, it achieves the goal and the
> continue could be removed.
>
> At the mean time, other two work in this patch:
> * the overlap check of two ranges could be done with one clause
> * one typo in comment is fixed.
>
> Signed-off-by: Wei Yang <[email protected]>

Acked-by: Tejun Heo <[email protected]>

Thanks.

--
tejun