NotificationsYou must be signed in to change notification settings
Fork5
Star26

Commit9e0bc7c

committed

Track spinlock delay in microsecond granularity.

On many platforms the OS will round the sleep time to millisecondresolution, but there is no reason for us to pre-emptively round theargument to pg_usleep.When the delay was measured in milliseconds and started from 1 ms, itsometimes took many attempts until the logic that increases the delay bymultiplying with a random value between 1 and 2 actually managed to bump itfrom 1 ms to 2 ms. That lead to a sequence of 1 ms waits until the delaystarted to increase. This wasn't really a problem but it looked odd if youobserved the waits. There is no measurable difference in performance, butit's more readable this way.Jeff Janes

1 parent9db4ad4 commit9e0bc7cCopy full SHA for 9e0bc7c

File tree

1 file changed

-11

lines changed

src/backend/storage/lmgr
- s_lock.c

1 file changed

-11

lines changed

`‎src/backend/storage/lmgr/s_lock.c`

Lines changed: 6 additions & 11 deletions

Original file line number	Diff line number	Diff line change
`@@ -81,17 +81,12 @@ s_lock(volatile slock_t lock, const char file, int line)`
`81`	`81`	`* so minutes.It seems better to fix the total number of tries (and thus`
`82`	`82`	`* the probability of unintended failure) than to fix the total time`
`83`	`83`	`* spent.`
`84`		`- *`
`85`		`- * The pg_usleep() delays are measured in milliseconds because 1 msec is a`
`86`		`- * common resolution limit at the OS level for newer platforms. On older`
`87`		`- * platforms the resolution limit is usually 10 msec, in which case the`
`88`		`- * total delay before timeout will be a bit more.`
`89`	`84`	`*/`
`90`	`85`	`#defineMIN_SPINS_PER_DELAY 10`
`91`	`86`	`#defineMAX_SPINS_PER_DELAY 1000`
`92`	`87`	`#defineNUM_DELAYS1000`
`93`		`-#defineMIN_DELAY_MSEC1`
`94`		`-#defineMAX_DELAY_MSEC1000`
	`88`	`+#defineMIN_DELAY_USEC1000L`
	`89`	`+#defineMAX_DELAY_USEC1000000L`
`95`	`90`
`96`	`91`	`intspins=0;`
`97`	`92`	`intdelays=0;`
`@@ -109,9 +104,9 @@ s_lock(volatile slock_t lock, const char file, int line)`
`109`	`104`	`s_lock_stuck(lock,file,line);`
`110`	`105`
`111`	`106`	`if (cur_delay==0)/* first time to delay? */`
`112`		`-cur_delay=MIN_DELAY_MSEC;`
	`107`	`+cur_delay=MIN_DELAY_USEC;`
`113`	`108`
`114`		`-pg_usleep(cur_delay*1000L);`
	`109`	`+pg_usleep(cur_delay);`
`115`	`110`
`116`	`111`	`#if defined(S_LOCK_TEST)`
`117`	`112`	`fprintf(stdout,"*");`
`@@ -122,8 +117,8 @@ s_lock(volatile slock_t lock, const char file, int line)`
`122`	`117`	`cur_delay+= (int) (cur_delay*`
`123`	`118`	`((double)random() / (double)MAX_RANDOM_VALUE)+0.5);`
`124`	`119`	`/* wrap back to minimum delay when max is exceeded */`
`125`		`-if (cur_delay>MAX_DELAY_MSEC)`
`126`		`-cur_delay=MIN_DELAY_MSEC;`
	`120`	`+if (cur_delay>MAX_DELAY_USEC)`
	`121`	`+cur_delay=MIN_DELAY_USEC;`
`127`	`122`
`128`	`123`	`spins=0;`
`129`	`124`	`}`

0 commit comments

Comments

(0)

Movatterモバイル変換

Navigation Menu

Search code, repositories, users, issues, pull requests...

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Commit9e0bc7c

File tree

1 file changed

1 file changed

`‎src/backend/storage/lmgr/s_lock.c`

0 commit comments