Improve spinlock performance for HP-UX, ia64, non-gcc.

robertmhaas · robertmhaas · commit c01c25fbe525 · 2011-08-29T10:05:48.000-04:00
At least on this architecture, it's very important to spin on a
non-atomic instruction and only retry the atomic once it appears
that it will succeed.  To fix this, split TAS() into two macros:
TAS(), for trying to grab the lock the first time, and TAS_SPIN(),
for spinning until we get it.  TAS_SPIN() defaults to same as TAS(),
but we can override it when we know there's a better way.

It's likely that some of the other cases in s_lock.h require
similar treatment, but this is the only one we've got conclusive
evidence for at present.
diff --git a/src/backend/storage/lmgr/s_lock.c b/src/backend/storage/lmgr/s_lock.c
@@ -96,7 +96,7 @@ s_lock(volatile slock_t *lock, const char *file, int line)
 	int			delays = 0;
 	int			cur_delay = 0;
 
-	while (TAS(lock))
+	while (TAS_SPIN(lock))
 	{
 		/* CPU-specific delay each time through the loop */
 		SPIN_DELAY();
diff --git a/src/include/storage/s_lock.h b/src/include/storage/s_lock.h
@@ -31,25 +31,33 @@
  *	macros at the bottom of the file.  Check if your platform can use
  *	these or needs to override them.
  *
- *  Usually, S_LOCK() is implemented in terms of an even lower-level macro
- *	TAS():
+ *  Usually, S_LOCK() is implemented in terms of even lower-level macros
+ *	TAS() and TAS_SPIN():
  *
  *	int TAS(slock_t *lock)
  *		Atomic test-and-set instruction.  Attempt to acquire the lock,
  *		but do *not* wait.	Returns 0 if successful, nonzero if unable
  *		to acquire the lock.
  *
- *	TAS() is NOT part of the API, and should never be called directly.
+ *	int TAS_SPIN(slock_t *lock)
+ *		Like TAS(), but this version is used when waiting for a lock
+ *		previously found to be contended.  Typically, this is the
+ *		same as TAS(), but on some architectures it's better to poll a
+ *		contended lock using an unlocked instruction and retry the
+ *		atomic test-and-set only when it appears free.
  *
- *	CAUTION: on some platforms TAS() may sometimes report failure to acquire
- *	a lock even when the lock is not locked.  For example, on Alpha TAS()
- *	will "fail" if interrupted.  Therefore TAS() should always be invoked
- *	in a retry loop, even if you are certain the lock is free.
+ *	TAS() and TAS_SPIN() are NOT part of the API, and should never be called
+ *	directly.
  *
- *	ANOTHER CAUTION: be sure that TAS() and S_UNLOCK() represent sequence
- *	points, ie, loads and stores of other values must not be moved across
- *	a lock or unlock.  In most cases it suffices to make the operation be
- *	done through a "volatile" pointer.
+ *	CAUTION: on some platforms TAS() and/or TAS_SPIN() may sometimes report
+ *	failure to acquire a lock even when the lock is not locked.  For example,
+ *	on Alpha TAS() will "fail" if interrupted.  Therefore a retry loop must
+ *	always be used, even if you are certain the lock is free.
+ *
+ *	ANOTHER CAUTION: be sure that TAS(), TAS_SPIN(), and S_UNLOCK() represent
+ *	sequence points, ie, loads and stores of other values must not be moved
+ *	across a lock or unlock.  In most cases it suffices to make the operation
+ *	be done through a "volatile" pointer.
  *
  *	On most supported platforms, TAS() uses a tas() function written
  *	in assembly language to execute a hardware atomic-test-and-set
@@ -727,6 +735,7 @@ typedef unsigned int slock_t;
 
 #include <ia64/sys/inline.h>
 #define TAS(lock) _Asm_xchg(_SZ_W, lock, 1, _LDHINT_NONE)
+#define TAS_SPIN(lock) (*(lock) ? 1 : TAS(lock))
 
 #endif	/* HPUX on IA64, non gcc */
 
@@ -925,6 +934,10 @@ extern int	tas(volatile slock_t *lock);		/* in port/.../tas.s, or
 #define TAS(lock)		tas(lock)
 #endif	 /* TAS */
 
+#if !defined(TAS_SPIN)
+#define TAS_SPIN(lock)	TAS(lock)
+#endif	 /* TAS_SPIN */
+
 
 /*
  * Platform-independent out-of-line support routines

Original file line number	Diff line number	Diff line change
`@@ -96,7 +96,7 @@ s_lock(volatile slock_t lock, const char file, int line)`
`96`	`96`	`int delays = 0;`
`97`	`97`	`int cur_delay = 0;`
`98`	`98`
`99`		`- while (TAS(lock))`
	`99`	`+ while (TAS_SPIN(lock))`
`100`	`100`	`{`
`101`	`101`	`/* CPU-specific delay each time through the loop */`
`102`	`102`	`SPIN_DELAY();`