memblock, numa: binary search node id
authorYinghai Lu <yinghai@kernel.org>
Wed, 11 Sep 2013 21:22:17 +0000 (14:22 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 11 Sep 2013 22:57:51 +0000 (15:57 -0700)
commite76b63f80d938a1319eb5fb0ae7ea69bddfbae38
tree4480ea31ebd4cbae35fcf7fa75c834ab06e39ffd
parent0bf598d863e3c741d47e3178d645f04c9d6c186c
memblock, numa: binary search node id

Current early_pfn_to_nid() on arch that support memblock go over
memblock.memory one by one, so will take too many try near the end.

We can use existing memblock_search to find the node id for given pfn,
that could save some time on bigger system that have many entries
memblock.memory array.

Here are the timing differences for several machines.  In each case with
the patch less time was spent in __early_pfn_to_nid().

                        3.11-rc5        with patch      difference (%)
                        --------        ----------      --------------
UV1: 256 nodes  9TB:     411.66          402.47         -9.19 (2.23%)
UV2: 255 nodes 16TB:    1141.02         1138.12         -2.90 (0.25%)
UV2:  64 nodes  2TB:     128.15          126.53         -1.62 (1.26%)
UV2:  32 nodes  2TB:     121.87          121.07         -0.80 (0.66%)
                        Time in seconds.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: Tejun Heo <tj@kernel.org>
Acked-by: Russ Anderson <rja@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
include/linux/memblock.h
mm/memblock.c
mm/page_alloc.c