[v2,for-next,0/4] optimise sbitmap deferred clear

Message

Pavel Begunkov Nov. 22, 2020, 3:35 p.m. UTC

sbitmap takes away some cycles for my tag-deficient test, removal of
locking in sbitmap_deferred_clear() gives +~1% throuhput.

[1/4] and [4/4] are simple, it'd be great if someone could double
check for ordering issues for other two patches.

v2: add 3rd (CAS -> atomic and) and 4th patches

Pavel Begunkov (4):
  sbitmap: optimise sbitmap_deferred_clear()
  sbitmap: remove swap_lock
  sbitmap: replace CAS with atomic and
  sbitmap: simplify wrap check

 include/linux/sbitmap.h |  5 -----
 lib/sbitmap.c           | 44 +++++++++++++++++------------------------
 2 files changed, 18 insertions(+), 31 deletions(-)

Comments

Jens Axboe Dec. 8, 2020, 12:13 a.m. UTC | #1

On 11/22/20 8:35 AM, Pavel Begunkov wrote:
> sbitmap takes away some cycles for my tag-deficient test, removal of
> locking in sbitmap_deferred_clear() gives +~1% throuhput.
> 
> [1/4] and [4/4] are simple, it'd be great if someone could double
> check for ordering issues for other two patches.
> 
> v2: add 3rd (CAS -> atomic and) and 4th patches

Applied, thanks.