Improve ExecStoreTuple to be smarter about replacing the contents of

a TupleTableSlot: instead of calling ExecClearTuple, inline the needed
operations, so that we can avoid redundant steps.  In particular, when
the old and new tuples are both on the same disk page, avoid releasing
and re-acquiring the buffer pin --- this saves work in both the bufmgr
and ResourceOwner modules.  To make this improvement actually useful,
partially revert a change I made on 2004-04-21 that caused SeqNext
et al to call ExecClearTuple before ExecStoreTuple.  The motivation
for that, to avoid grabbing the BufMgrLock separately for releasing
the old buffer and grabbing the new one, no longer applies.  My
profiling says that this saves about 5% of the CPU time for an
all-in-memory seqscan.

1 files changed, 2 insertions, 11 deletions

diff --git a/src/backend/executor/nodeIndexscan.c b/src/backend/executor/nodeIndexscan.c
index 6e639502c1d..4f6fadfde49 100644
--- a/src/backend/executor/nodeIndexscan.c
+++ b/src/backend/executor/nodeIndexscan.c
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/executor/nodeIndexscan.c,v 1.105 2005/11/22 18:17:10 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/executor/nodeIndexscan.c,v 1.106 2005/11/25 04:24:48 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -75,15 +75,6 @@ IndexNext(IndexScanState *node)
 	scanrelid = ((IndexScan *) node->ss.ps.plan)->scan.scanrelid;
 
 	/*
-	 * Clear any reference to the previously returned tuple.  The idea here is
-	 * to not have the tuple slot be the last holder of a pin on that tuple's
-	 * buffer; if it is, we'll need a separate visit to the bufmgr to release
-	 * the buffer.	By clearing here, we get to have the release done by
-	 * ReleaseAndReadBuffer inside index_getnext.
-	 */
-	ExecClearTuple(slot);
-
-	/*
 	 * Check if we are evaluating PlanQual for tuple of this relation.
 	 * Additional checking is not good, but no other way for now. We could
 	 * introduce new nodes for this case and handle IndexScan --> NewNode
@@ -93,7 +84,7 @@ IndexNext(IndexScanState *node)
 		estate->es_evTuple[scanrelid - 1] != NULL)
 	{
 		if (estate->es_evTupleNull[scanrelid - 1])
-			return slot;		/* return empty slot */
+			return ExecClearTuple(slot);
 
 		ExecStoreTuple(estate->es_evTuple[scanrelid - 1],
 					   slot, InvalidBuffer, false);