]> err.no Git - linux-2.6/commit
[PATCH] sched: likely profiling
authorNick Piggin <npiggin@suse.de>
Wed, 11 Oct 2006 08:21:52 +0000 (01:21 -0700)
committerLinus Torvalds <torvalds@g5.osdl.org>
Wed, 11 Oct 2006 18:14:22 +0000 (11:14 -0700)
commitbeed33a816204cb402c69266475b6a60a2433ceb
tree4eaa7e5a1ccf2960d1478774cdfcab671384accb
parentf33d9bd50478c9a969b65f58feb6b69a3ad478cb
[PATCH] sched: likely profiling

This likely profiling is pretty fun. I found a few possible problems
in sched.c.

This patch may be not measurable, but when I did measure long ago,
nooping (un)likely cost a couple of % on scheduler heavy benchmarks, so
it all adds up.

Tweak some branch hints:

- the 2nd 64 bits in the bitmask is likely to be populated, because it
  contains the first 28 bits (nearly 3/4) of the normal priorities.
  (ratio of 669669:691 ~= 1000:1).

- it isn't unlikely that context switching switches to another process. it
  might be very rapidly switching to and from the idle process (ratio of
  475815:419004 and 471330:423544). Let the branch predictor decide.

- preempt_enable seems to be very often called in a nested preempt_disable
  or with interrupts disabled (ratio of 3567760:87965 ~= 40:1)

Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Cc: Daniel Walker <dwalker@mvista.com>
Cc: Hua Zhong <hzhong@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
include/asm-generic/bitops/sched.h
kernel/sched.c