Re: [PATCH 0/5] branch hint tweaks

[Brian Gerst]

On Tue, Nov 24, 2009 at 5:54 AM, Tim Blechmann <tim@xxxxxxxxxx> wrote:
> hi all,
>
> doing some branch hint profiling of 2.6.31 on my nehalem machine showed
> a few
> incorrect branch hints.
> these patches remove or change the branch hints
>
> Tim Blechmann (5):
> Âprocess_64: remove branch hint
> Âsched.c: change branch hint
> Âslab.c: remove branch hint
> Âsched_fair.c: remove branch hint
> Âworkqueue.c: remove branch hint
>
> Âarch/x86/kernel/process_64.c | Â Â4 ++--
> Âkernel/sched.c        |  Â4 ++--
> Âkernel/sched_fair.c     Â|  Â2 +-
> Âkernel/workqueue.c      |  Â2 +-
> Âmm/slab.c          Â|  Â2 +-
> Â5 files changed, 7 insertions(+), 7 deletions(-)

Did you run profiling tests again after making these changes to see if
they had any effect?  likely() and unlikely() are only hints.  GCC
doesn't have to follow them, or it could be broken in recent GCC
versions.

I'm not sure what version of GCC added this, but I wonder if this
option will fix the problem:

-fno-guess-branch-probability
    Do not guess branch probabilities using heuristics.

    GCC will use heuristics to guess branch probabilities if they are
not provided by profiling feedback (-fprofile-arcs). These heuristics
are based on the control flow graph. If some branch probabilities are
specified by `__builtin_expect', then the heuristics will be used to
guess branch probabilities for the rest of the control flow graph,
taking the `__builtin_expect' info into account. The interactions
between the heuristics and `__builtin_expect' can be complex, and in
some cases, it may be useful to disable the heuristics so that the
effects of `__builtin_expect' are easier to understand.

    The default is -fguess-branch-probability at levels -O, -O2, -O3, -Os.

--
Brian Gerst
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/