mm/page_alloc: inline __rmqueue_pcplist
authorJesper Dangaard Brouer <brouer@redhat.com>
Fri, 30 Apr 2021 06:01:55 +0000 (23:01 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 30 Apr 2021 18:20:43 +0000 (11:20 -0700)
When __alloc_pages_bulk() got introduced two callers of __rmqueue_pcplist
exist and the compiler chooses to not inline this function.

  ./scripts/bloat-o-meter vmlinux-before vmlinux-inline__rmqueue_pcplist
  add/remove: 0/1 grow/shrink: 2/0 up/down: 164/-125 (39)
  Function                                     old     new   delta
  rmqueue                                     2197    2296     +99
  __alloc_pages_bulk                          1921    1986     +65
  __rmqueue_pcplist                            125       -    -125
  Total: Before=19374127, After=19374166, chg +0.00%

modprobe page_bench04_bulk loops=$((10**7))

Type:time_bulk_page_alloc_free_array
 -  Per elem: 106 cycles(tsc) 29.595 ns (step:64)
 - (measurement period time:0.295955434 sec time_interval:295955434)
 - (invoke count:10000000 tsc_interval:1065447105)

Before:
 - Per elem: 110 cycles(tsc) 30.633 ns (step:64)

Link: https://lkml.kernel.org/r/20210325114228.27719-6-mgorman@techsingularity.net
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Reviewed-by: Alexander Lobakin <alobakin@pm.me>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Alexander Duyck <alexander.duyck@gmail.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Chuck Lever <chuck.lever@oracle.com>
Cc: David Miller <davem@davemloft.net>
Cc: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/page_alloc.c

index 3f2291695b19ec33e313a183abeccb5c91e84231..91455f0d22c12e3e155b2529633ac0a7bfb47a1f 100644 (file)
@@ -3456,7 +3456,8 @@ static inline void zone_statistics(struct zone *preferred_zone, struct zone *z)
 }
 
 /* Remove page from the per-cpu list, caller must protect the list */
-static struct page *__rmqueue_pcplist(struct zone *zone, int migratetype,
+static inline
+struct page *__rmqueue_pcplist(struct zone *zone, int migratetype,
                        unsigned int alloc_flags,
                        struct per_cpu_pages *pcp,
                        struct list_head *list)