[PATCH] perf_events: improve task_sched_in()

[PATCH] perf_events: improve task_sched_in()

[eranian]

This patch is an optimization in perf_event_task_sched_in() to avoid scheduling
the events twice in a row. Without it, the perf_disable()/perf_enable() pair
is invoked twice, thereby pinned events counts while scheduling flexible events
and we go throuh hw_perf_enable() twice. By encapsulating, the whole sequence
into perf_disable()/perf_enable() we ensure, hw_perf_enable() is going to be
invoked only once because of the refcount protection.

	Signed-off-by: Stephane Eranian <eranian@google.com>
--
 perf_event.c |    4 ++++
 1 file changed, 4 insertions(+)

--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -1382,6 +1382,8 @@ void perf_event_task_sched_in(struct task_struct *task)
 	if (cpuctx->task_ctx == ctx)
 		return;
 
+	perf_disable();
+
 	/*
 	 * We want to keep the following priority order:
 	 * cpu pinned (that don't need to move), task pinned,
@@ -1394,6 +1396,8 @@ void perf_event_task_sched_in(struct task_struct *task)
 	ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE);
 
 	cpuctx->task_ctx = ctx;
+
+	perf_enable();
 }
 
 #define MAX_INTERRUPTS (~0ULL)

Re: [PATCH] perf_events: improve task_sched_in()

[Peter Zijlstra]

On Wed, 2010-03-10 at 22:26 -0800, eranian@google.com wrote:
> This patch is an optimization in perf_event_task_sched_in() to avoid scheduling
> the events twice in a row. Without it, the perf_disable()/perf_enable() pair
> is invoked twice, thereby pinned events counts while scheduling flexible events
> and we go throuh hw_perf_enable() twice. By encapsulating, the whole sequence
> into perf_disable()/perf_enable() we ensure, hw_perf_enable() is going to be
> invoked only once because of the refcount protection.

Agreed, this makes perfect sense.

Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>

> Signed-off-by: Stephane Eranian <eranian@google.com>
> --
>  perf_event.c |    4 ++++
>  1 file changed, 4 insertions(+)
> 
> --- a/kernel/perf_event.c
> +++ b/kernel/perf_event.c
> @@ -1382,6 +1382,8 @@ void perf_event_task_sched_in(struct task_struct *task)
>  	if (cpuctx->task_ctx == ctx)
>  		return;
>  
> +	perf_disable();
> +
>  	/*
>  	 * We want to keep the following priority order:
>  	 * cpu pinned (that don't need to move), task pinned,
> @@ -1394,6 +1396,8 @@ void perf_event_task_sched_in(struct task_struct *task)
>  	ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE);
>  
>  	cpuctx->task_ctx = ctx;
> +
> +	perf_enable();
>  }
>  
>  #define MAX_INTERRUPTS (~0ULL)

[tip:perf/core] perf_events: Improve task_sched_in()

[tip-bot for eranian@google.com]

Commit-ID:  9b33fa6ba0e2f90fdf407501db801c2511121564
Gitweb:     http://git.kernel.org/tip/9b33fa6ba0e2f90fdf407501db801c2511121564
Author:     eranian@google.com <eranian@google.com>
AuthorDate: Wed, 10 Mar 2010 22:26:05 -0800
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Thu, 11 Mar 2010 15:23:28 +0100

perf_events: Improve task_sched_in()

This patch is an optimization in perf_event_task_sched_in() to avoid
scheduling the events twice in a row.

Without it, the perf_disable()/perf_enable() pair is invoked twice,
thereby pinned events counts while scheduling flexible events and we go
throuh hw_perf_enable() twice.

By encapsulating, the whole sequence into perf_disable()/perf_enable() we
ensure, hw_perf_enable() is going to be invoked only once because of the
refcount protection.

Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1268288765-5326-1-git-send-email-eranian@google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 kernel/perf_event.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index 52c69a3..3853d49 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -1368,6 +1368,8 @@ void perf_event_task_sched_in(struct task_struct *task)
 	if (cpuctx->task_ctx == ctx)
 		return;
 
+	perf_disable();
+
 	/*
 	 * We want to keep the following priority order:
 	 * cpu pinned (that don't need to move), task pinned,
@@ -1380,6 +1382,8 @@ void perf_event_task_sched_in(struct task_struct *task)
 	ctx_sched_in(ctx, cpuctx, EVENT_FLEXIBLE);
 
 	cpuctx->task_ctx = ctx;
+
+	perf_enable();
 }
 
 #define MAX_INTERRUPTS (~0ULL)