Slightly reduce BufMgrLock contention

This patch prevents btbulkdelete() from calling WriteNoReleaseBuffer()
several times for the same buffer.  Thus it saves a few
LWLockAquire(BufMgrLock, LW_EXCLUSIVE) and LWLockRelease(BufMgrLock)
calls.

Maybe we do not need a BufMgrLock at all, because we have a super
exclusive lock on the buffer?  I was not sure and decided to stay on
the safe side ...

Servus
 Manfred

diff -ruN ../base/src/backend/access/nbtree/nbtree.c 
src/backend/access/nbtree/nbtree.c
--- ../base/src/backend/access/nbtree/nbtree.c  2002-06-21 02:12:14.000000000 
+0200
+++ src/backend/access/nbtree/nbtree.c  2002-08-30 11:44:35.000000000 +0200
@@ -615,6 +615,7 @@
        {
                Buffer          buf;
                BlockNumber lockedBlock = InvalidBlockNumber;
+               bool            dirty = false;
 
                /* we have the buffer pinned and locked */
                buf = so->btso_curbuf;
@@ -662,14 +663,19 @@
                                        LockBuffer(buf, BUFFER_LOCK_UNLOCK);
                                        LockBufferForCleanup(buf);
                                        lockedBlock = blkno;
+                                       dirty = false;
                                }
                                else
                                {
                                        /* Okay to delete the item from the 
page */
                                        _bt_itemdel(rel, buf, current);
 
-                                       /* Mark buffer dirty, but keep the lock 
and pin */
-                                       WriteNoReleaseBuffer(buf);
+                                       if (!dirty)
+                                       {
+                                               /* Mark buffer dirty, but keep 
the lock and pin */
+                                               WriteNoReleaseBuffer(buf);
+                                               dirty = true;
+                                       }
 
                                        tuples_removed += 1;
                                }

---------------------------(end of broadcast)---------------------------
TIP 5: Have you checked our extensive FAQ?

http://www.postgresql.org/users-lounge/docs/faq.html