Skip to contentSkip to conversation

cfq-iosched: simplify prio-unboost code

PATCH In review Corrado Zoccolo 2 participants 4 messages Oct 26, 2009 – Oct 29, 2009 Block layer

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 cfq-iosched: simplify prio-unboost code

Eliminate redundant checks.

---
 block/cfq-iosched.c |    8 +++-----
 1 files changed, 3 insertions(+), 5 deletions(-)
block/cfq-iosched.c +3 -5
index 6e9b395..244bc8a 100644
@@ -2611,12 +2611,10 @@ static void cfq_prio_boost(struct cfq_queue *cfqq)
 			cfqq->ioprio = IOPRIO_NORM;
 	} else {
 		/*
-		 * check if we need to unboost the queue
+		 * unboost the queue (if needed)
 		 */
-		if (cfqq->ioprio_class != cfqq->org_ioprio_class)
-			cfqq->ioprio_class = cfqq->org_ioprio_class;
-		if (cfqq->ioprio != cfqq->org_ioprio)
-			cfqq->ioprio = cfqq->org_ioprio;
+		cfqq->ioprio_class = cfqq->org_ioprio_class;
+		cfqq->ioprio = cfqq->org_ioprio;
 	}
 }
 
Signature · 1 line
1.6.2.5
#1 lore raw
Quoted 15 lines · Corrado Zoccolo

Eliminate redundant checks.

---
 block/cfq-iosched.c |    8 +++-----
 1 files changed, 3 insertions(+), 5 deletions(-)
block/cfq-iosched.c +0 -0
index 6e9b395..244bc8a 100644
@@ -2611,12 +2611,10 @@ static void cfq_prio_boost(struct cfq_queue *cfqq)
 			cfqq->ioprio = IOPRIO_NORM;
 	} else {
 		/*

block/cfq-iosched.c static void cfq_prio_boost(struct cfq_queue *cfqq) · line 2618

Quoted 10 lines
block/cfq-iosched.c +3 -5
-		 * check if we need to unboost the queue
+		 * unboost the queue (if needed)
 		 */
-		if (cfqq->ioprio_class != cfqq->org_ioprio_class)
-			cfqq->ioprio_class = cfqq->org_ioprio_class;
-		if (cfqq->ioprio != cfqq->org_ioprio)
-			cfqq->ioprio = cfqq->org_ioprio;
+		cfqq->ioprio_class = cfqq->org_ioprio_class;
+		cfqq->ioprio = cfqq->org_ioprio;
 	}

Not sure I see much gain in that, the previous code makes it explicit that it may not be different and avoid dirtying cfqq if it isn't.

Signature · 1 line
Jens Axboe
#2 lore raw
Quoted 30 lines · Jens Axboe
Quoted 15 lines · Corrado Zoccolo

Eliminate redundant checks.

---
 block/cfq-iosched.c |    8 +++-----

 1 files changed, 3 insertions(+), 5 deletions(-)
block/cfq-iosched.c +0 -0
index 6e9b395..244bc8a 100644
@@ -2611,12 +2611,10 @@ static void cfq_prio_boost(struct cfq_queue *cfqq)
                       cfqq->ioprio = IOPRIO_NORM;
       } else {
               /*

block/cfq-iosched.c static void cfq_prio_boost(struct cfq_queue *cfqq) · line 2618

Quoted 10 lines
block/cfq-iosched.c +3 -5
-              * check if we need to unboost the queue
+              * unboost the queue (if needed)
                */
-             if (cfqq->ioprio_class != cfqq->org_ioprio_class)
-                     cfqq->ioprio_class = cfqq->org_ioprio_class;
-             if (cfqq->ioprio != cfqq->org_ioprio)
-                     cfqq->ioprio = cfqq->org_ioprio;
+             cfqq->ioprio_class = cfqq->org_ioprio_class;
+             cfqq->ioprio = cfqq->org_ioprio;
       }

Not sure I see much gain in that, the previous code makes it explicit that it may not be different and avoid dirtying cfqq if it isn't.

Quoted 1 line

From readability p.o.v., it is simpler to understand the direct assignment.

The comment says that they might already be equal, but the code states clearly what is the final state, i.e. they will surely be equal at the end-

Moreover, using two branches to avoid dirtying a cache line is not a win performance-wise, given the penalty on mispredicted branches, and the fact that the interval between two executions of this code is large, so the branch predictor entries for those would be already replaced by userspace ones.

If on some exotic MP hardware dirtying a cache line could be very expensive, a similar check could be done in hardware, and it would be more efficient, in fact it could distinguish the 2 cases:
* we already have the shared cache line locally -> we can compare the value and absorb the write if it doesn't change
* we don't have the cache line locally -> we acquire the cache line in exclusive mode, so we can do the write directly (on the contrary, the previous code would first acquire the cache line as shared, and if necessary, upgrade it to exclusive, possibly increasing the traffic).

Corrado

Quoted 5 lines

--
Jens Axboe

#3 lore raw
Quoted 31 lines · Corrado Zoccolo
Quoted 21 lines · Jens Axboe
Quoted 20 lines · Corrado Zoccolo

Eliminate redundant checks.

---
 block/cfq-iosched.c |    8 +++-----

 1 files changed, 3 insertions(+), 5 deletions(-)
block/cfq-iosched.c +1 -3
index 6e9b395..244bc8a 100644
@@ -2611,12 +2611,10 @@ static void cfq_prio_boost(struct cfq_queue *cfqq)
                       cfqq->ioprio = IOPRIO_NORM;
       } else {
               /*
-              * check if we need to unboost the queue
+              * unboost the queue (if needed)
                */
-             if (cfqq->ioprio_class != cfqq->org_ioprio_class)
-                     cfqq->ioprio_class = cfqq->org_ioprio_class;
Quoted 8 lines

block/cfq-iosched.c static void cfq_prio_boost(struct cfq_queue *cfqq) · line 2618

Quoted 5 lines
block/cfq-iosched.c +2 -2
-             if (cfqq->ioprio != cfqq->org_ioprio)
-                     cfqq->ioprio = cfqq->org_ioprio;
+             cfqq->ioprio_class = cfqq->org_ioprio_class;
+             cfqq->ioprio = cfqq->org_ioprio;
       }

Not sure I see much gain in that, the previous code makes it explicit that it may not be different and avoid dirtying cfqq if it isn't.

Quoted 1 line

From readability p.o.v., it is simpler to understand the direct assignment.

Debatable :-)

Quoted 3 lines

The comment says that they might already be equal, but the code states clearly what is the final state, i.e. they will surely be equal at the end-

Naturally that is true, since we assign if they aren't equal.

Quoted 6 lines

Moreover, using two branches to avoid dirtying a cache line is not a win performance-wise, given the penalty on mispredicted branches, and the fact that the interval between two executions of this code is large, so the branch predictor entries for those would be already replaced by userspace ones.

Quoted 9 lines

If on some exotic MP hardware dirtying a cache line could be very expensive, a similar check could be done in hardware, and it would be more efficient, in fact it could distinguish the 2 cases:
* we already have the shared cache line locally -> we can compare the value and absorb the write if it doesn't change
* we don't have the cache line locally -> we acquire the cache line in exclusive mode, so we can do the write directly (on the contrary, the previous code would first acquire the cache line as shared, and if necessary, upgrade it to exclusive, possibly increasing the traffic).

If we just improve the comment to state that they may or may not be equal before assigning them, then I'm fine with changing it. Code readability to me trumps micro optimizations, because frankly it's not like either case would make any measurable difference. Comments aren't quite as good as code, but it's good enough here at least.

Signature · 1 line
Jens Axboe