[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <20240601025536.25682-1-rgbi3307@naver.com>
Date: Sat, 1 Jun 2024 11:55:36 +0900
From: Jung-JaeJoon <rgbi3307@...il.com>
To: "Liam R. Howlett" <Liam.Howlett@...cle.com>
Cc: Jung-JaeJoon <rgbi3307@...il.com>,
maple-tree@...ts.infradead.org,
linux-mm@...ck.org,
linux-kernel@...r.kernel.org
Subject: [PATCH] maple_tree: add mas_node_count() before going to slow_path in mas_wr_modify()
From: Jung-JaeJoon <rgbi3307@...il.com>
If there are not enough nodes, mas_node_count() set an error state via mas_set_err()
and return control flow to the beginning.
In the return flow, mas_nomem() checks the error status, allocates new nodes,
and resumes execution again.
In particular,
if this happens in mas_split() in the slow_path section executed in mas_wr_modify(),
unnecessary work is repeated, causing a slowdown in speed as below flow:
_begin:
mas_wr_modify() --> if (new_end >= mt_slots[wr_mas->type]) --> goto slow_path
slow_path:
--> mas_wr_bnode() --> mas_store_b_node() --> mas_commit_b_node() --> mas_split()
--> mas_node_count() return to _begin
But, in the above flow, if mas_node_count() is executed before entering slow_path,
execution efficiency is improved by allocating nodes without entering slow_path repeatedly.
Signed-off-by: JaeJoon Jung <rgbi3307@...il.com>
---
lib/maple_tree.c | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/lib/maple_tree.c b/lib/maple_tree.c
index 2d7d27e6ae3c..8ffabd73619f 100644
--- a/lib/maple_tree.c
+++ b/lib/maple_tree.c
@@ -4176,8 +4176,13 @@ static inline void mas_wr_modify(struct ma_wr_state *wr_mas)
* path.
*/
new_end = mas_wr_new_end(wr_mas);
- if (new_end >= mt_slots[wr_mas->type])
+ if (new_end >= mt_slots[wr_mas->type]) {
+ mas->depth = mas_mt_height(mas);
+ mas_node_count(mas, 1 + mas->depth * 2);
+ if (mas_is_err(mas))
+ return;
goto slow_path;
+ }
/* Attempt to append */
if (mas_wr_append(wr_mas, new_end))
--
2.17.1
Powered by blists - more mailing lists