Improve performance of subsystems on top of SLRU

More precisely, what we do here is make the SLRU cache sizes
configurable with new GUCs, so that sites with high concurrency and big
ranges of transactions in flight (resp. multixacts/subtransactions) can
benefit from bigger caches.  In order for this to work with good
performance, two additional changes are made:

1. the cache is divided in "banks" (to borrow terminology from CPU
   caches), and algorithms such as eviction buffer search only affect
   one specific bank.  This forestalls the problem that linear searching
   for a specific buffer across the whole cache takes too long: we only
   have to search the specific bank, whose size is small.  This work is
   authored by Andrey Borodin.

2. Change the locking regime for the SLRU banks, so that each bank uses
   a separate LWLock.  This allows for increased scalability.  This work
   is authored by Dilip Kumar.  (A part of this was previously committed as
   d172b717c6f4.)

Special care is taken so that the algorithms that can potentially
traverse more than one bank release one bank's lock before acquiring the
next.  This should happen rarely, but particularly clog.c's group commit
feature needed code adjustment to cope with this.  I (Álvaro) also added
lots of comments to make sure the design is sound.

The new GUCs match the names introduced by bcdfa5f2e2 in the
pg_stat_slru view.

The default values for these parameters are similar to the previous
sizes of each SLRU.  commit_ts, clog and subtrans accept value 0, which
means to adjust by dividing shared_buffers by 512 (so 2MB for every 1GB
of shared_buffers), with a cap of 8MB.  (A new slru.c function
SimpleLruAutotuneBuffers() was added to support this.)  The cap was
previously 1MB for clog, so for sites with more than 512MB of shared
memory the total memory used increases, which is likely a good tradeoff.
However, other SLRUs (notably multixact ones) retain smaller sizes and
don't support a configured value of 0.  These values based on
shared_buffers may need to be revisited, but that's an easy change.

There was some resistance to adding these new GUCs: it would be better
to adjust to memory pressure automatically somehow, for example by
stealing memory from shared_buffers (where the caches can grow and
shrink naturally).  However, doing that seems to be a much larger
project and one which has made virtually no progress in several years,
and because this is such a pain point for so many users, here we take
the pragmatic approach.

Author: Andrey Borodin <x4mmm@yandex-team.ru>
Author: Dilip Kumar <dilipbalaut@gmail.com>
Reviewed-by: Amul Sul, Gilles Darold, Anastasia Lubennikova,
	Ivan Lazarev, Robert Haas, Thomas Munro, Tomas Vondra,
	Yura Sokolov, Васильев Дмитрий (Dmitry Vasiliev).
Discussion: https://postgr.es/m/2BEC2B3F-9B61-4C1D-9FB5-5FAB0F05EF86@yandex-team.ru
Discussion: https://postgr.es/m/CAFiTN-vzDvNz=ExGXz6gdyjtzGixKSqs0mKHMmaQ8sOSEFZ33A@mail.gmail.com

src/backend/storage/lmgr/lwlock.c

@@ -163,6 +163,13 @@ static const char *const BuiltinTrancheNames[] = {
 	[LWTRANCHE_LAUNCHER_HASH] = "LogicalRepLauncherHash",
 	[LWTRANCHE_DSM_REGISTRY_DSA] = "DSMRegistryDSA",
 	[LWTRANCHE_DSM_REGISTRY_HASH] = "DSMRegistryHash",
+	[LWTRANCHE_COMMITTS_SLRU] = "CommitTSSLRU",
+	[LWTRANCHE_MULTIXACTOFFSET_SLRU] = "MultixactOffsetSLRU",
+	[LWTRANCHE_MULTIXACTMEMBER_SLRU] = "MultixactMemberSLRU",
+	[LWTRANCHE_NOTIFY_SLRU] = "NotifySLRU",
+	[LWTRANCHE_SERIAL_SLRU] = "SerialSLRU",
+	[LWTRANCHE_SUBTRANS_SLRU] = "SubtransSLRU",
+	[LWTRANCHE_XACT_SLRU] = "XactSLRU",
 };
 
 StaticAssertDecl(lengthof(BuiltinTrancheNames) ==
@@ -776,7 +783,7 @@ GetLWLockIdentifier(uint32 classId, uint16 eventId)
  * in mode.
  *
  * This function will not block waiting for a lock to become free - that's the
- * callers job.
+ * caller's job.
  *
  * Returns true if the lock isn't free and we need to wait.
  */

src/backend/storage/lmgr/lwlocknames.txt

@@ -16,11 +16,11 @@ WALBufMappingLock					7
 WALWriteLock						8
 ControlFileLock						9
 # 10 was CheckpointLock
-XactSLRULock						11
-SubtransSLRULock					12
+# 11 was XactSLRULock
+# 12 was SubtransSLRULock
 MultiXactGenLock					13
-MultiXactOffsetSLRULock				14
-MultiXactMemberSLRULock				15
+# 14 was MultiXactOffsetSLRULock
+# 15 was MultiXactMemberSLRULock
 RelCacheInitLock					16
 CheckpointerCommLock				17
 TwoPhaseStateLock					18
@@ -31,19 +31,19 @@ AutovacuumLock						22
 AutovacuumScheduleLock				23
 SyncScanLock						24
 RelationMappingLock					25
-NotifySLRULock						26
+#26 was NotifySLRULock
 NotifyQueueLock						27
 SerializableXactHashLock			28
 SerializableFinishedListLock		29
 SerializablePredicateListLock		30
-SerialSLRULock						31
+# 31 was SerialSLRULock
 SyncRepLock							32
 BackgroundWorkerLock				33
 DynamicSharedMemoryControlLock		34
 AutoFileLock						35
 ReplicationSlotAllocationLock		36
 ReplicationSlotControlLock			37
-CommitTsSLRULock					38
+#38 was CommitTsSLRULock
 CommitTsLock						39
 ReplicationOriginLock				40
 MultiXactTruncationLock				41

src/backend/storage/lmgr/predicate.c

@@ -213,6 +213,7 @@
 #include "storage/predicate_internals.h"
 #include "storage/proc.h"
 #include "storage/procarray.h"
+#include "utils/guc_hooks.h"
 #include "utils/rel.h"
 #include "utils/snapmgr.h"
 
@@ -813,9 +814,9 @@ SerialInit(void)
 	 */
 	SerialSlruCtl->PagePrecedes = SerialPagePrecedesLogically;
 	SimpleLruInit(SerialSlruCtl, "serializable",
-				  NUM_SERIAL_BUFFERS, 0, SerialSLRULock, "pg_serial",
-				  LWTRANCHE_SERIAL_BUFFER, SYNC_HANDLER_NONE,
-				  false);
+				  serializable_buffers, 0, "pg_serial",
+				  LWTRANCHE_SERIAL_BUFFER, LWTRANCHE_SERIAL_SLRU,
+				  SYNC_HANDLER_NONE, false);
 #ifdef USE_ASSERT_CHECKING
 	SerialPagePrecedesLogicallyUnitTests();
 #endif
@@ -841,6 +842,15 @@ SerialInit(void)
 	}
 }
 
+/*
+ * GUC check_hook for serializable_buffers
+ */
+bool
+check_serial_buffers(int *newval, void **extra, GucSource source)
+{
+	return check_slru_buffers("serializable_buffers", newval);
+}
+
 /*
  * Record a committed read write serializable xid and the minimum
  * commitSeqNo of any transactions to which this xid had a rw-conflict out.
@@ -854,15 +864,17 @@ SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo)
 	int			slotno;
 	int64		firstZeroPage;
 	bool		isNewPage;
+	LWLock	   *lock;
 
 	Assert(TransactionIdIsValid(xid));
 
 	targetPage = SerialPage(xid);
+	lock = SimpleLruGetBankLock(SerialSlruCtl, targetPage);
 
 	/*
-	 * In this routine, we must hold both SerialControlLock and SerialSLRULock
-	 * simultaneously while making the SLRU data catch up with the new state
-	 * that we determine.
+	 * In this routine, we must hold both SerialControlLock and the SLRU bank
+	 * lock simultaneously while making the SLRU data catch up with the new
+	 * state that we determine.
 	 */
 	LWLockAcquire(SerialControlLock, LW_EXCLUSIVE);
 
@@ -898,7 +910,7 @@ SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo)
 	if (isNewPage)
 		serialControl->headPage = targetPage;
 
-	LWLockAcquire(SerialSLRULock, LW_EXCLUSIVE);
+	LWLockAcquire(lock, LW_EXCLUSIVE);
 
 	if (isNewPage)
 	{
@@ -916,7 +928,7 @@ SerialAdd(TransactionId xid, SerCommitSeqNo minConflictCommitSeqNo)
 	SerialValue(slotno, xid) = minConflictCommitSeqNo;
 	SerialSlruCtl->shared->page_dirty[slotno] = true;
 
-	LWLockRelease(SerialSLRULock);
+	LWLockRelease(lock);
 	LWLockRelease(SerialControlLock);
 }
 
@@ -950,13 +962,13 @@ SerialGetMinConflictCommitSeqNo(TransactionId xid)
 		return 0;
 
 	/*
-	 * The following function must be called without holding SerialSLRULock,
+	 * The following function must be called without holding SLRU bank lock,
 	 * but will return with that lock held, which must then be released.
 	 */
 	slotno = SimpleLruReadPage_ReadOnly(SerialSlruCtl,
 										SerialPage(xid), xid);
 	val = SerialValue(slotno, xid);
-	LWLockRelease(SerialSLRULock);
+	LWLockRelease(SimpleLruGetBankLock(SerialSlruCtl, SerialPage(xid)));
 	return val;
 }
 
@@ -1367,7 +1379,7 @@ PredicateLockShmemSize(void)
 
 	/* Shared memory structures for SLRU tracking of old committed xids. */
 	size = add_size(size, sizeof(SerialControlData));
-	size = add_size(size, SimpleLruShmemSize(NUM_SERIAL_BUFFERS, 0));
+	size = add_size(size, SimpleLruShmemSize(serializable_buffers, 0));
 
 	return size;
 }