From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
To: linux-kernel@vger.kernel.org
Cc: mingo@elte.hu, laijs@cn.fujitsu.com, dipankar@in.ibm.com,
akpm@linux-foundation.org, mathieu.desnoyers@polymtl.ca,
josh@joshtriplett.org, niv@us.ibm.com, tglx@linutronix.de,
peterz@infradead.org, rostedt@goodmis.org,
Valdis.Kletnieks@vt.edu, dhowells@redhat.com,
eric.dumazet@gmail.com, darren@dvhart.com,
"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Subject: [PATCH RFC tip/core/rcu 17/20] rcu: limit rcu_node leaf-level fanout
Date: Fri, 17 Dec 2010 12:54:48 -0800 [thread overview]
Message-ID: <1292619291-2468-17-git-send-email-paulmck@linux.vnet.ibm.com> (raw)
In-Reply-To: <20101217205433.GA10199@linux.vnet.ibm.com>
Some recent benchmarks have indicated possible lock contention on the
leaf-level rcu_node locks. This commit therefore limits the number of
CPUs per leaf-level rcu_node structure to 16, in other words, there
can be at most 16 rcu_data structures fanning into a given rcu_node
structure. Prior to this, the limit was 32 on 32-bit systems and 64 on
64-bit systems.
Note that the fanout of non-leaf rcu_node structures is unchanged. The
organization of accesses to the rcu_node tree is such that references
to non-leaf rcu_node structures are much less frequent than to the
leaf structures.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
kernel/rcutree.c | 3 ++-
kernel/rcutree.h | 43 ++++++++++++++++++++++++-------------------
2 files changed, 26 insertions(+), 20 deletions(-)
diff --git a/kernel/rcutree.c b/kernel/rcutree.c
index c39ec5b..01c8ad3 100644
--- a/kernel/rcutree.c
+++ b/kernel/rcutree.c
@@ -1869,8 +1869,9 @@ static void __init rcu_init_levelspread(struct rcu_state *rsp)
{
int i;
- for (i = NUM_RCU_LVLS - 1; i >= 0; i--)
+ for (i = NUM_RCU_LVLS - 1; i > 0; i--)
rsp->levelspread[i] = CONFIG_RCU_FANOUT;
+ rsp->levelspread[0] = RCU_FANOUT_LEAF;
}
#else /* #ifdef CONFIG_RCU_FANOUT_EXACT */
static void __init rcu_init_levelspread(struct rcu_state *rsp)
diff --git a/kernel/rcutree.h b/kernel/rcutree.h
index 1a54be2..e8f057e 100644
--- a/kernel/rcutree.h
+++ b/kernel/rcutree.h
@@ -31,46 +31,51 @@
/*
* Define shape of hierarchy based on NR_CPUS and CONFIG_RCU_FANOUT.
* In theory, it should be possible to add more levels straightforwardly.
- * In practice, this has not been tested, so there is probably some
- * bug somewhere.
+ * In practice, this did work well going from three levels to four.
+ * Of course, your mileage may vary.
*/
#define MAX_RCU_LVLS 4
-#define RCU_FANOUT (CONFIG_RCU_FANOUT)
-#define RCU_FANOUT_SQ (RCU_FANOUT * RCU_FANOUT)
-#define RCU_FANOUT_CUBE (RCU_FANOUT_SQ * RCU_FANOUT)
-#define RCU_FANOUT_FOURTH (RCU_FANOUT_CUBE * RCU_FANOUT)
-
-#if NR_CPUS <= RCU_FANOUT
+#if CONFIG_RCU_FANOUT > 16
+#define RCU_FANOUT_LEAF 16
+#else /* #if CONFIG_RCU_FANOUT > 16 */
+#define RCU_FANOUT_LEAF (CONFIG_RCU_FANOUT)
+#endif /* #else #if CONFIG_RCU_FANOUT > 16 */
+#define RCU_FANOUT_1 (RCU_FANOUT_LEAF)
+#define RCU_FANOUT_2 (RCU_FANOUT_1 * CONFIG_RCU_FANOUT)
+#define RCU_FANOUT_3 (RCU_FANOUT_2 * CONFIG_RCU_FANOUT)
+#define RCU_FANOUT_4 (RCU_FANOUT_3 * CONFIG_RCU_FANOUT)
+
+#if NR_CPUS <= RCU_FANOUT_1
# define NUM_RCU_LVLS 1
# define NUM_RCU_LVL_0 1
# define NUM_RCU_LVL_1 (NR_CPUS)
# define NUM_RCU_LVL_2 0
# define NUM_RCU_LVL_3 0
# define NUM_RCU_LVL_4 0
-#elif NR_CPUS <= RCU_FANOUT_SQ
+#elif NR_CPUS <= RCU_FANOUT_2
# define NUM_RCU_LVLS 2
# define NUM_RCU_LVL_0 1
-# define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT)
+# define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_1)
# define NUM_RCU_LVL_2 (NR_CPUS)
# define NUM_RCU_LVL_3 0
# define NUM_RCU_LVL_4 0
-#elif NR_CPUS <= RCU_FANOUT_CUBE
+#elif NR_CPUS <= RCU_FANOUT_3
# define NUM_RCU_LVLS 3
# define NUM_RCU_LVL_0 1
-# define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_SQ)
-# define NUM_RCU_LVL_2 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT)
-# define NUM_RCU_LVL_3 NR_CPUS
+# define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_2)
+# define NUM_RCU_LVL_2 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_1)
+# define NUM_RCU_LVL_3 (NR_CPUS)
# define NUM_RCU_LVL_4 0
-#elif NR_CPUS <= RCU_FANOUT_FOURTH
+#elif NR_CPUS <= RCU_FANOUT_4
# define NUM_RCU_LVLS 4
# define NUM_RCU_LVL_0 1
-# define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_CUBE)
-# define NUM_RCU_LVL_2 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_SQ)
-# define NUM_RCU_LVL_3 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT)
-# define NUM_RCU_LVL_4 NR_CPUS
+# define NUM_RCU_LVL_1 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_3)
+# define NUM_RCU_LVL_2 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_2)
+# define NUM_RCU_LVL_3 DIV_ROUND_UP(NR_CPUS, RCU_FANOUT_1)
+# define NUM_RCU_LVL_4 (NR_CPUS)
#else
# error "CONFIG_RCU_FANOUT insufficient for NR_CPUS"
-#endif /* #if (NR_CPUS) <= RCU_FANOUT */
+#endif /* #if (NR_CPUS) <= RCU_FANOUT_1 */
#define RCU_SUM (NUM_RCU_LVL_0 + NUM_RCU_LVL_1 + NUM_RCU_LVL_2 + NUM_RCU_LVL_3 + NUM_RCU_LVL_4)
#define NUM_RCU_NODES (RCU_SUM - NR_CPUS)
--
1.7.3.2
next prev parent reply other threads:[~2010-12-17 20:58 UTC|newest]
Thread overview: 34+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-12-17 20:54 [PATCH tip/core/rcu 0/20] second preview of RCU patches for 2.6.38 Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 01/20] rcu: add priority-inversion testing to rcutorture Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 02/20] rcu: move TINY_RCU from softirq to kthread Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 03/20] rcu: priority boosting for TINY_PREEMPT_RCU Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 04/20] rcu: add tracing for TINY_RCU and TINY_PREEMPT_RCU Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 05/20] rcu: document TINY_RCU and TINY_PREEMPT_RCU tracing Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 06/20] rcu: Distinguish between boosting and boosted Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 07/20] rcu: get rid of obsolete "classic" names in TREE_RCU tracing Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 08/20] rcu,cleanup: move synchronize_sched_expedited() out of sched.c Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 09/20] rcu,cleanup: simplify the code when cpu is dying Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 10/20] rcu: update documentation/comments for Lai's adoption patch Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 11/20] rcu: fix race condition in synchronize_sched_expedited() Paul E. McKenney
2010-12-18 15:52 ` Tejun Heo
2010-12-18 19:58 ` Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 12/20] rcu: Make synchronize_srcu_expedited() fast if running readers Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 13/20] rcu: increase synchronize_sched_expedited() batching Paul E. McKenney
2010-12-18 16:13 ` Tejun Heo
2010-12-18 20:14 ` Paul E. McKenney
2010-12-19 9:43 ` Tejun Heo
2010-12-19 16:35 ` Paul E. McKenney
2010-12-20 10:33 ` Peter Zijlstra
2010-12-20 13:40 ` Mathieu Desnoyers
2010-12-20 10:31 ` Peter Zijlstra
2010-12-21 7:58 ` Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 14/20] rcu: Stop chasing QS if another CPU did it for us Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 15/20] rcu: Keep gpnum and completed fields synchronized Paul E. McKenney
2010-12-20 2:13 ` Lai Jiangshan
2010-12-20 2:14 ` Frederic Weisbecker
2010-12-20 16:51 ` Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 16/20] rcu: fine-tune grace-period begin/end checks Paul E. McKenney
2010-12-17 20:54 ` Paul E. McKenney [this message]
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 18/20] rcu: reduce __call_rcu()-induced contention on rcu_node structures Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 19/20] rculist: fix borked __list_for_each_rcu() macro Paul E. McKenney
2010-12-17 20:54 ` [PATCH RFC tip/core/rcu 20/20] rcu: remove unused " Paul E. McKenney
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1292619291-2468-17-git-send-email-paulmck@linux.vnet.ibm.com \
--to=paulmck@linux.vnet.ibm.com \
--cc=Valdis.Kletnieks@vt.edu \
--cc=akpm@linux-foundation.org \
--cc=darren@dvhart.com \
--cc=dhowells@redhat.com \
--cc=dipankar@in.ibm.com \
--cc=eric.dumazet@gmail.com \
--cc=josh@joshtriplett.org \
--cc=laijs@cn.fujitsu.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mathieu.desnoyers@polymtl.ca \
--cc=mingo@elte.hu \
--cc=niv@us.ibm.com \
--cc=peterz@infradead.org \
--cc=rostedt@goodmis.org \
--cc=tglx@linutronix.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®