[PATCHv2,0/3] zsmalloc: small compaction improvements

Message

Sergey Senozhatsky June 24, 2023, 5:12 a.m. UTC

Hi,
	A tiny series that can reduce the number of
find_alloced_obj() invocations (which perform a linear
scan of sub-page) during compaction. Inspired by Alexey
Romanov's findings.

v2:
-- picked up a patch from Minchan

Minchan Kim (1):
  zsmalloc: remove zs_compact_control

Sergey Senozhatsky (2):
  zsmalloc: do not scan for allocated objects in empty zspage
  zsmalloc: move migration destination zspage inuse check

 mm/zsmalloc.c | 50 ++++++++++++++++++++------------------------------
 1 file changed, 20 insertions(+), 30 deletions(-)

Comments

Sergey Senozhatsky July 5, 2023, 1:28 p.m. UTC | #1

On (23/06/24 14:12), Sergey Senozhatsky wrote:
> Hi,
> 	A tiny series that can reduce the number of
> find_alloced_obj() invocations (which perform a linear
> scan of sub-page) during compaction. Inspired by Alexey
> Romanov's findings.
> 
> v2:
> -- picked up a patch from Minchan
> 
> Minchan Kim (1):
>   zsmalloc: remove zs_compact_control
> 
> Sergey Senozhatsky (2):
>   zsmalloc: do not scan for allocated objects in empty zspage
>   zsmalloc: move migration destination zspage inuse check

Just for the record,

A synthetic zsmalloc fragmentation+compaction test (100% reproducible)

           num find_tagged_obj() calls     num iterations in find_tagged_obj()

base                 545699                              812899
patch #1             460701                              691821
patch #2             422372                              651372

// lower is better


patch #1 is "zsmalloc: do not scan for allocated objects in empty zspage"
patch #2 is "zsmalloc: move migration destination zspage inuse check"