perf: Add unlikely() to the ring-buffer code

Add unlikely() annotations to 'slow' paths:

When having a sampling event but no output buffer; you have bigger
issues -- also the bail is still faster than actually doing the work.

When having a sampling event but a control page only buffer, you have
bigger issues -- again the bail is still faster than actually doing
work.

Optimize for the case where you're not loosing events -- again, not
doing the work is still faster but make sure that when you have to
actually do work its as fast as possible.

The typical watermark is 1/2 the buffer size, so most events will not
take this path.

Shrinks perf_output_begin() by 16 bytes on x86_64-defconfig.

Signed-off-by: Peter Zijlstra <[email protected]>
Cc: Benjamin Herrenschmidt <[email protected]>
Cc: Frederic Weisbecker <[email protected]>
Cc: Mathieu Desnoyers <[email protected]>
Cc: Michael Ellerman <[email protected]>
Cc: Michael Neuling <[email protected]>
Cc: "Paul E. McKenney" <[email protected]>
Cc: [email protected]
Cc: Vince Weaver <[email protected]>
Cc: Victor Kaplansky <[email protected]>
Cc: Oleg Nesterov <[email protected]>
Cc: Anton Blanchard <[email protected]>
Link: http://lkml.kernel.org/n/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>

diff --git a/kernel/events/ring_buffer.c b/kernel/events/ring_buffer.c
index 6929c5848d4ff5ad19a20d6804d809d45da82396..383cde47617636377ce7f0267259884557c29f2d 100644 (file)
--- a/kernel/events/ring_buffer.c
+++ b/kernel/events/ring_buffer.c
@@ -121,17 +121,17 @@ int perf_output_begin(struct perf_output_handle *handle,
                event = event->parent;
 
        rb = rcu_dereference(event->rb);
-       if (!rb)
+       if (unlikely(!rb))
                goto out;
 
-       handle->rb      = rb;
-       handle->event   = event;
-
-       if (!rb->nr_pages)
+       if (unlikely(!rb->nr_pages))
                goto out;
 
+       handle->rb    = rb;
+       handle->event = event;
+
        have_lost = local_read(&rb->lost);
-       if (have_lost) {
+       if (unlikely(have_lost)) {
                lost_event.header.size = sizeof(lost_event);
                perf_event_header__init_id(&lost_event.header, &sample_data,
                                           event);
@@ -157,7 +157,7 @@ int perf_output_begin(struct perf_output_handle *handle,
                head += size;
        } while (local_cmpxchg(&rb->head, offset, head) != offset);
 
-       if (head - local_read(&rb->wakeup) > rb->watermark)
+       if (unlikely(head - local_read(&rb->wakeup) > rb->watermark))
                local_add(rb->watermark, &rb->wakeup);
 
        handle->page = offset >> (PAGE_SHIFT + page_order(rb));
@@ -167,7 +167,7 @@ int perf_output_begin(struct perf_output_handle *handle,
        handle->addr += handle->size;
        handle->size = (PAGE_SIZE << page_order(rb)) - handle->size;
 
-       if (have_lost) {
+       if (unlikely(have_lost)) {
                lost_event.header.type = PERF_RECORD_LOST;
                lost_event.header.misc = 0;
                lost_event.id          = event->id;