[PATCH 1/5] ring-buffer: Make write slow path out of line

[Steven Rostedt]

From: Steven Rostedt <srostedt@xxxxxxxxxx>

Gcc inlines the slow path of the ring buffer write which can
hurt performance. This patch simply forces the slow path function
rb_move_tail() to always be a function.

The ring_buffer_benchmark module with reader_disabled=1 shows that
this patch changes the time to record an event from 135 ns to
132 ns. (3 ns or 2.22% improvement)

Signed-off-by: Steven Rostedt <rostedt@xxxxxxxxxxx>
---
 kernel/trace/ring_buffer.c |    7 +++++--
 1 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
index bca9637..0b88df8 100644
--- a/kernel/trace/ring_buffer.c
+++ b/kernel/trace/ring_buffer.c
@@ -1823,7 +1823,10 @@ rb_reset_tail(struct ring_buffer_per_cpu *cpu_buffer,
 	local_sub(length, &tail_page->write);
 }
 
-static struct ring_buffer_event *
+/*
+ * This is the slow path, force gcc not to inline it.
+ */
+static noinline struct ring_buffer_event *
 rb_move_tail(struct ring_buffer_per_cpu *cpu_buffer,
 	     unsigned long length, unsigned long tail,
 	     struct buffer_page *tail_page, u64 *ts)
@@ -1943,7 +1946,7 @@ __rb_reserve_next(struct ring_buffer_per_cpu *cpu_buffer,
 	tail = write - length;
 
 	/* See if we shot pass the end of this buffer page */
-	if (write > BUF_PAGE_SIZE)
+	if (unlikely(write > BUF_PAGE_SIZE))
 		return rb_move_tail(cpu_buffer, length, tail,
 				    tail_page, ts);
 
-- 
1.7.1


--
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/