Add optimizer and executor support for parallel index scans.

In combination with 569174f1be92be93f5366212cc46960d28a5c5cd, which
taught the btree AM how to perform parallel index scans, this allows
parallel index scan plans on btree indexes.  This infrastructure
should be general enough to support parallel index scans for other
index AMs as well, if someone updates them to support parallel
scans.

Amit Kapila, reviewed and tested by Anastasia Lubennikova, Tushar
Ahuja, and Haribabu Kommi, and me.

15 files changed, 297 insertions, 44 deletions

diff --git a/src/backend/access/brin/brin.c b/src/backend/access/brin/brin.c
index 4ff046b4b0..b22563bf7c 100644
--- a/src/backend/access/brin/brin.c
+++ b/src/backend/access/brin/brin.c
@@ -93,6 +93,7 @@ brinhandler(PG_FUNCTION_ARGS)
 	amroutine->amstorage = true;
 	amroutine->amclusterable = false;
 	amroutine->ampredlocks = false;
+	amroutine->amcanparallel = false;
 	amroutine->amkeytype = InvalidOid;
 
 	amroutine->ambuild = brinbuild;
diff --git a/src/backend/access/gin/ginutil.c b/src/backend/access/gin/ginutil.c
index a98d4fc397..d03d59da6a 100644
--- a/src/backend/access/gin/ginutil.c
+++ b/src/backend/access/gin/ginutil.c
@@ -50,6 +50,7 @@ ginhandler(PG_FUNCTION_ARGS)
 	amroutine->amstorage = true;
 	amroutine->amclusterable = false;
 	amroutine->ampredlocks = false;
+	amroutine->amcanparallel = false;
 	amroutine->amkeytype = InvalidOid;
 
 	amroutine->ambuild = ginbuild;
diff --git a/src/backend/access/gist/gist.c b/src/backend/access/gist/gist.c
index 96ead531ea..6593771361 100644
--- a/src/backend/access/gist/gist.c
+++ b/src/backend/access/gist/gist.c
@@ -71,6 +71,7 @@ gisthandler(PG_FUNCTION_ARGS)
 	amroutine->amstorage = true;
 	amroutine->amclusterable = true;
 	amroutine->ampredlocks = false;
+	amroutine->amcanparallel = false;
 	amroutine->amkeytype = InvalidOid;
 
 	amroutine->ambuild = gistbuild;
diff --git a/src/backend/access/hash/hash.c b/src/backend/access/hash/hash.c
index bca77a80c3..24510e78f5 100644
--- a/src/backend/access/hash/hash.c
+++ b/src/backend/access/hash/hash.c
@@ -67,6 +67,7 @@ hashhandler(PG_FUNCTION_ARGS)
 	amroutine->amstorage = false;
 	amroutine->amclusterable = false;
 	amroutine->ampredlocks = false;
+	amroutine->amcanparallel = false;
 	amroutine->amkeytype = INT4OID;
 
 	amroutine->ambuild = hashbuild;
diff --git a/src/backend/access/nbtree/nbtree.c b/src/backend/access/nbtree/nbtree.c
index cbc575d5cf..775f2ff1f8 100644
--- a/src/backend/access/nbtree/nbtree.c
+++ b/src/backend/access/nbtree/nbtree.c
@@ -140,6 +140,7 @@ bthandler(PG_FUNCTION_ARGS)
 	amroutine->amstorage = false;
 	amroutine->amclusterable = true;
 	amroutine->ampredlocks = true;
+	amroutine->amcanparallel = true;
 	amroutine->amkeytype = InvalidOid;
 
 	amroutine->ambuild = btbuild;
diff --git a/src/backend/access/spgist/spgutils.c b/src/backend/access/spgist/spgutils.c
index 78846bec66..e57ac49c6b 100644
--- a/src/backend/access/spgist/spgutils.c
+++ b/src/backend/access/spgist/spgutils.c
@@ -49,6 +49,7 @@ spghandler(PG_FUNCTION_ARGS)
 	amroutine->amstorage = false;
 	amroutine->amclusterable = false;
 	amroutine->ampredlocks = false;
+	amroutine->amcanparallel = false;
 	amroutine->amkeytype = InvalidOid;
 
 	amroutine->ambuild = spgbuild;
diff --git a/src/backend/executor/execParallel.c b/src/backend/executor/execParallel.c
index 784dbaf590..98d4f1eca7 100644
--- a/src/backend/executor/execParallel.c
+++ b/src/backend/executor/execParallel.c
@@ -28,6 +28,7 @@
 #include "executor/nodeCustom.h"
 #include "executor/nodeForeignscan.h"
 #include "executor/nodeSeqscan.h"
+#include "executor/nodeIndexscan.h"
 #include "executor/tqueue.h"
 #include "nodes/nodeFuncs.h"
 #include "optimizer/planmain.h"
@@ -197,6 +198,10 @@ ExecParallelEstimate(PlanState *planstate, ExecParallelEstimateContext *e)
 				ExecSeqScanEstimate((SeqScanState *) planstate,
 									e->pcxt);
 				break;
+			case T_IndexScanState:
+				ExecIndexScanEstimate((IndexScanState *) planstate,
+									  e->pcxt);
+				break;
 			case T_ForeignScanState:
 				ExecForeignScanEstimate((ForeignScanState *) planstate,
 										e->pcxt);
@@ -249,6 +254,10 @@ ExecParallelInitializeDSM(PlanState *planstate,
 				ExecSeqScanInitializeDSM((SeqScanState *) planstate,
 										 d->pcxt);
 				break;
+			case T_IndexScanState:
+				ExecIndexScanInitializeDSM((IndexScanState *) planstate,
+										   d->pcxt);
+				break;
 			case T_ForeignScanState:
 				ExecForeignScanInitializeDSM((ForeignScanState *) planstate,
 											 d->pcxt);
@@ -725,6 +734,9 @@ ExecParallelInitializeWorker(PlanState *planstate, shm_toc *toc)
 			case T_SeqScanState:
 				ExecSeqScanInitializeWorker((SeqScanState *) planstate, toc);
 				break;
+			case T_IndexScanState:
+				ExecIndexScanInitializeWorker((IndexScanState *) planstate, toc);
+				break;
 			case T_ForeignScanState:
 				ExecForeignScanInitializeWorker((ForeignScanState *) planstate,
 												toc);
diff --git a/src/backend/executor/nodeIndexscan.c b/src/backend/executor/nodeIndexscan.c
index 5734550d2c..0a9dfdbaf3 100644
--- a/src/backend/executor/nodeIndexscan.c
+++ b/src/backend/executor/nodeIndexscan.c
@@ -22,6 +22,9 @@
  *		ExecEndIndexScan		releases all storage.
  *		ExecIndexMarkPos		marks scan position.
  *		ExecIndexRestrPos		restores scan position.
+ *		ExecIndexScanEstimate	estimates DSM space needed for parallel index scan
+ *		ExecIndexScanInitializeDSM initialize DSM for parallel indexscan
+ *		ExecIndexScanInitializeWorker attach to DSM info in parallel worker
  */
 #include "postgres.h"
 
@@ -514,6 +517,18 @@ ExecIndexScan(IndexScanState *node)
 void
 ExecReScanIndexScan(IndexScanState *node)
 {
+	bool		reset_parallel_scan = true;
+
+	/*
+	 * If we are here to just update the scan keys, then don't reset parallel
+	 * scan.  We don't want each of the participating process in the parallel
+	 * scan to update the shared parallel scan state at the start of the scan.
+	 * It is quite possible that one of the participants has already begun
+	 * scanning the index when another has yet to start it.
+	 */
+	if (node->iss_NumRuntimeKeys != 0 && !node->iss_RuntimeKeysReady)
+		reset_parallel_scan = false;
+
 	/*
 	 * If we are doing runtime key calculations (ie, any of the index key
 	 * values weren't simple Consts), compute the new key values.  But first,
@@ -539,10 +554,21 @@ ExecReScanIndexScan(IndexScanState *node)
 			reorderqueue_pop(node);
 	}
 
-	/* reset index scan */
-	index_rescan(node->iss_ScanDesc,
-				 node->iss_ScanKeys, node->iss_NumScanKeys,
-				 node->iss_OrderByKeys, node->iss_NumOrderByKeys);
+	/*
+	 * Reset (parallel) index scan.  For parallel-aware nodes, the scan
+	 * descriptor is initialized during actual execution of node and we can
+	 * reach here before that (ex. during execution of nest loop join).  So,
+	 * avoid updating the scan descriptor at that time.
+	 */
+	if (node->iss_ScanDesc)
+	{
+		index_rescan(node->iss_ScanDesc,
+					 node->iss_ScanKeys, node->iss_NumScanKeys,
+					 node->iss_OrderByKeys, node->iss_NumOrderByKeys);
+
+		if (reset_parallel_scan && node->iss_ScanDesc->parallel_scan)
+			index_parallelrescan(node->iss_ScanDesc);
+	}
 	node->iss_ReachedEnd = false;
 
 	ExecScanReScan(&node->ss);
@@ -1013,22 +1039,29 @@ ExecInitIndexScan(IndexScan *node, EState *estate, int eflags)
 	}
 
 	/*
-	 * Initialize scan descriptor.
+	 * for parallel-aware node, we initialize the scan descriptor after
+	 * initializing the shared memory for parallel execution.
 	 */
-	indexstate->iss_ScanDesc = index_beginscan(currentRelation,
-											   indexstate->iss_RelationDesc,
-											   estate->es_snapshot,
-											   indexstate->iss_NumScanKeys,
+	if (!node->scan.plan.parallel_aware)
+	{
+		/*
+		 * Initialize scan descriptor.
+		 */
+		indexstate->iss_ScanDesc = index_beginscan(currentRelation,
+												indexstate->iss_RelationDesc,
+												   estate->es_snapshot,
+												 indexstate->iss_NumScanKeys,
 											 indexstate->iss_NumOrderByKeys);
 
-	/*
-	 * If no run-time keys to calculate, go ahead and pass the scankeys to the
-	 * index AM.
-	 */
-	if (indexstate->iss_NumRuntimeKeys == 0)
-		index_rescan(indexstate->iss_ScanDesc,
-					 indexstate->iss_ScanKeys, indexstate->iss_NumScanKeys,
+		/*
+		 * If no run-time keys to calculate, go ahead and pass the scankeys to
+		 * the index AM.
+		 */
+		if (indexstate->iss_NumRuntimeKeys == 0)
+			index_rescan(indexstate->iss_ScanDesc,
+					   indexstate->iss_ScanKeys, indexstate->iss_NumScanKeys,
 				indexstate->iss_OrderByKeys, indexstate->iss_NumOrderByKeys);
+	}
 
 	/*
 	 * all done.
@@ -1590,3 +1623,91 @@ ExecIndexBuildScanKeys(PlanState *planstate, Relation index,
 	else if (n_array_keys != 0)
 		elog(ERROR, "ScalarArrayOpExpr index qual found where not allowed");
 }
+
+/* ----------------------------------------------------------------
+ *						Parallel Scan Support
+ * ----------------------------------------------------------------
+ */
+
+/* ----------------------------------------------------------------
+ *		ExecIndexScanEstimate
+ *
+ *		estimates the space required to serialize indexscan node.
+ * ----------------------------------------------------------------
+ */
+void
+ExecIndexScanEstimate(IndexScanState *node,
+					  ParallelContext *pcxt)
+{
+	EState	   *estate = node->ss.ps.state;
+
+	node->iss_PscanLen = index_parallelscan_estimate(node->iss_RelationDesc,
+													 estate->es_snapshot);
+	shm_toc_estimate_chunk(&pcxt->estimator, node->iss_PscanLen);
+	shm_toc_estimate_keys(&pcxt->estimator, 1);
+}
+
+/* ----------------------------------------------------------------
+ *		ExecIndexScanInitializeDSM
+ *
+ *		Set up a parallel index scan descriptor.
+ * ----------------------------------------------------------------
+ */
+void
+ExecIndexScanInitializeDSM(IndexScanState *node,
+						   ParallelContext *pcxt)
+{
+	EState	   *estate = node->ss.ps.state;
+	ParallelIndexScanDesc piscan;
+
+	piscan = shm_toc_allocate(pcxt->toc, node->iss_PscanLen);
+	index_parallelscan_initialize(node->ss.ss_currentRelation,
+								  node->iss_RelationDesc,
+								  estate->es_snapshot,
+								  piscan);
+	shm_toc_insert(pcxt->toc, node->ss.ps.plan->plan_node_id, piscan);
+	node->iss_ScanDesc =
+		index_beginscan_parallel(node->ss.ss_currentRelation,
+								 node->iss_RelationDesc,
+								 node->iss_NumScanKeys,
+								 node->iss_NumOrderByKeys,
+								 piscan);
+
+	/*
+	 * If no run-time keys to calculate, go ahead and pass the scankeys to the
+	 * index AM.
+	 */
+	if (node->iss_NumRuntimeKeys == 0)
+		index_rescan(node->iss_ScanDesc,
+					 node->iss_ScanKeys, node->iss_NumScanKeys,
+					 node->iss_OrderByKeys, node->iss_NumOrderByKeys);
+}
+
+/* ----------------------------------------------------------------
+ *		ExecIndexScanInitializeWorker
+ *
+ *		Copy relevant information from TOC into planstate.
+ * ----------------------------------------------------------------
+ */
+void
+ExecIndexScanInitializeWorker(IndexScanState *node, shm_toc *toc)
+{
+	ParallelIndexScanDesc piscan;
+
+	piscan = shm_toc_lookup(toc, node->ss.ps.plan->plan_node_id);
+	node->iss_ScanDesc =
+		index_beginscan_parallel(node->ss.ss_currentRelation,
+								 node->iss_RelationDesc,
+								 node->iss_NumScanKeys,
+								 node->iss_NumOrderByKeys,
+								 piscan);
+
+	/*
+	 * If no run-time keys to calculate, go ahead and pass the scankeys to the
+	 * index AM.
+	 */
+	if (node->iss_NumRuntimeKeys == 0)
+		index_rescan(node->iss_ScanDesc,
+					 node->iss_ScanKeys, node->iss_NumScanKeys,
+					 node->iss_OrderByKeys, node->iss_NumOrderByKeys);
+}
diff --git a/src/backend/optimizer/path/allpaths.c b/src/backend/optimizer/path/allpaths.c
index 85505c57d3..eeacf815e3 100644
--- a/src/backend/optimizer/path/allpaths.c
+++ b/src/backend/optimizer/path/allpaths.c
@@ -127,8 +127,6 @@ static void subquery_push_qual(Query *subquery,
 static void recurse_push_qual(Node *setOp, Query *topquery,
 				  RangeTblEntry *rte, Index rti, Node *qual);
 static void remove_unused_subquery_outputs(Query *subquery, RelOptInfo *rel);
-static int compute_parallel_worker(RelOptInfo *rel, BlockNumber heap_pages,
-						BlockNumber index_pages);
 
 
 /*
@@ -2885,7 +2883,7 @@ remove_unused_subquery_outputs(Query *subquery, RelOptInfo *rel)
  * "heap_pages" is the number of pages from the table that we expect to scan.
  * "index_pages" is the number of pages from the index that we expect to scan.
  */
-static int
+int
 compute_parallel_worker(RelOptInfo *rel, BlockNumber heap_pages,
 						BlockNumber index_pages)
 {
diff --git a/src/backend/optimizer/path/costsize.c b/src/backend/optimizer/path/costsize.c
index a43daa744c..d01630f8db 100644
--- a/src/backend/optimizer/path/costsize.c
+++ b/src/backend/optimizer/path/costsize.c
@@ -391,7 +391,8 @@ cost_gather(GatherPath *path, PlannerInfo *root,
  * we have to fetch from the table, so they don't reduce the scan cost.
  */
 void
-cost_index(IndexPath *path, PlannerInfo *root, double loop_count)
+cost_index(IndexPath *path, PlannerInfo *root, double loop_count,
+		   bool partial_path)
 {
 	IndexOptInfo *index = path->indexinfo;
 	RelOptInfo *baserel = index->rel;
@@ -400,6 +401,7 @@ cost_index(IndexPath *path, PlannerInfo *root, double loop_count)
 	List	   *qpquals;
 	Cost		startup_cost = 0;
 	Cost		run_cost = 0;
+	Cost		cpu_run_cost = 0;
 	Cost		indexStartupCost;
 	Cost		indexTotalCost;
 	Selectivity indexSelectivity;
@@ -413,6 +415,8 @@ cost_index(IndexPath *path, PlannerInfo *root, double loop_count)
 	Cost		cpu_per_tuple;
 	double		tuples_fetched;
 	double		pages_fetched;
+	double		rand_heap_pages;
+	double		index_pages;
 
 	/* Should only be applied to base relations */
 	Assert(IsA(baserel, RelOptInfo) &&
@@ -459,7 +463,8 @@ cost_index(IndexPath *path, PlannerInfo *root, double loop_count)
 	amcostestimate = (amcostestimate_function) index->amcostestimate;
 	amcostestimate(root, path, loop_count,
 				   &indexStartupCost, &indexTotalCost,
-				   &indexSelectivity, &indexCorrelation);
+				   &indexSelectivity, &indexCorrelation,
+				   &index_pages);
 
 	/*
 	 * Save amcostestimate's results for possible use in bitmap scan planning.
@@ -526,6 +531,8 @@ cost_index(IndexPath *path, PlannerInfo *root, double loop_count)
 		if (indexonly)
 			pages_fetched = ceil(pages_fetched * (1.0 - baserel->allvisfrac));
 
+		rand_heap_pages = pages_fetched;
+
 		max_IO_cost = (pages_fetched * spc_random_page_cost) / loop_count;
 
 		/*
@@ -564,6 +571,8 @@ cost_index(IndexPath *path, PlannerInfo *root, double loop_count)
 		if (indexonly)
 			pages_fetched = ceil(pages_fetched * (1.0 - baserel->allvisfrac));
 
+		rand_heap_pages = pages_fetched;
+
 		/* max_IO_cost is for the perfectly uncorrelated case (csquared=0) */
 		max_IO_cost = pages_fetched * spc_random_page_cost;
 
@@ -583,6 +592,29 @@ cost_index(IndexPath *path, PlannerInfo *root, double loop_count)
 			min_IO_cost = 0;
 	}
 
+	if (partial_path)
+	{
+		/*
+		 * Estimate the number of parallel workers required to scan index. Use
+		 * the number of heap pages computed considering heap fetches won't be
+		 * sequential as for parallel scans the pages are accessed in random
+		 * order.
+		 */
+		path->path.parallel_workers = compute_parallel_worker(baserel,
+											   (BlockNumber) rand_heap_pages,
+												  (BlockNumber) index_pages);
+
+		/*
+		 * Fall out if workers can't be assigned for parallel scan, because in
+		 * such a case this path will be rejected.  So there is no benefit in
+		 * doing extra computation.
+		 */
+		if (path->path.parallel_workers <= 0)
+			return;
+
+		path->path.parallel_aware = true;
+	}
+
 	/*
 	 * Now interpolate based on estimated index order correlation to get total
 	 * disk I/O cost for main table accesses.
@@ -602,11 +634,24 @@ cost_index(IndexPath *path, PlannerInfo *root, double loop_count)
 	startup_cost += qpqual_cost.startup;
 	cpu_per_tuple = cpu_tuple_cost + qpqual_cost.per_tuple;
 
-	run_cost += cpu_per_tuple * tuples_fetched;
+	cpu_run_cost += cpu_per_tuple * tuples_fetched;
 
 	/* tlist eval costs are paid per output row, not per tuple scanned */
 	startup_cost += path->path.pathtarget->cost.startup;
-	run_cost += path->path.pathtarget->cost.per_tuple * path->path.rows;
+	cpu_run_cost += path->path.pathtarget->cost.per_tuple * path->path.rows;
+
+	/* Adjust costing for parallelism, if used. */
+	if (path->path.parallel_workers > 0)
+	{
+		double		parallel_divisor = get_parallel_divisor(&path->path);
+
+		path->path.rows = clamp_row_est(path->path.rows / parallel_divisor);
+
+		/* The CPU cost is divided among all the workers. */
+		cpu_run_cost /= parallel_divisor;
+	}
+
+	run_cost += cpu_run_cost;
 
 	path->path.startup_cost = startup_cost;
 	path->path.total_cost = startup_cost + run_cost;
diff --git a/src/backend/optimizer/path/indxpath.c b/src/backend/optimizer/path/indxpath.c
index 5283468988..56eccafd7b 100644
--- a/src/backend/optimizer/path/indxpath.c
+++ b/src/backend/optimizer/path/indxpath.c
@@ -813,7 +813,7 @@ get_index_paths(PlannerInfo *root, RelOptInfo *rel,
 /*
  * build_index_paths
  *	  Given an index and a set of index clauses for it, construct zero
- *	  or more IndexPaths.
+ *	  or more IndexPaths. It also constructs zero or more partial IndexPaths.
  *
  * We return a list of paths because (1) this routine checks some cases
  * that should cause us to not generate any IndexPath, and (2) in some
@@ -1042,8 +1042,41 @@ build_index_paths(PlannerInfo *root, RelOptInfo *rel,
 								  NoMovementScanDirection,
 								  index_only_scan,
 								  outer_relids,
-								  loop_count);
+								  loop_count,
+								  false);
 		result = lappend(result, ipath);
+
+		/*
+		 * If appropriate, consider parallel index scan.  We don't allow
+		 * parallel index scan for bitmap or index only scans.
+		 */
+		if (index->amcanparallel && !index_only_scan &&
+			rel->consider_parallel && outer_relids == NULL &&
+			scantype != ST_BITMAPSCAN)
+		{
+			ipath = create_index_path(root, index,
+									  index_clauses,
+									  clause_columns,
+									  orderbyclauses,
+									  orderbyclausecols,
+									  useful_pathkeys,
+									  index_is_ordered ?
+									  ForwardScanDirection :
+									  NoMovementScanDirection,
+									  index_only_scan,
+									  outer_relids,
+									  loop_count,
+									  true);
+
+			/*
+			 * if, after costing the path, we find that it's not worth
+			 * using parallel workers, just free it.
+			 */
+			if (ipath->path.parallel_workers > 0)
+				add_partial_path(rel, (Path *) ipath);
+			else
+				pfree(ipath);
+		}
 	}
 
 	/*
@@ -1066,8 +1099,36 @@ build_index_paths(PlannerInfo *root, RelOptInfo *rel,
 									  BackwardScanDirection,
 									  index_only_scan,
 									  outer_relids,
-									  loop_count);
+									  loop_count,
+									  false);
 			result = lappend(result, ipath);
+
+			/* If appropriate, consider parallel index scan */
+			if (index->amcanparallel && !index_only_scan &&
+				rel->consider_parallel && outer_relids == NULL &&
+				scantype != ST_BITMAPSCAN)
+			{
+				ipath = create_index_path(root, index,
+										  index_clauses,
+										  clause_columns,
+										  NIL,
+										  NIL,
+										  useful_pathkeys,
+										  BackwardScanDirection,
+										  index_only_scan,
+										  outer_relids,
+										  loop_count,
+										  true);
+
+				/*
+				 * if, after costing the path, we find that it's not worth
+				 * using parallel workers, just free it.
+				 */
+				if (ipath->path.parallel_workers > 0)
+					add_partial_path(rel, (Path *) ipath);
+				else
+					pfree(ipath);
+			}
 		}
 	}
 
diff --git a/src/backend/optimizer/plan/planner.c b/src/backend/optimizer/plan/planner.c
index abb4f12cea..3d33d46971 100644
--- a/src/backend/optimizer/plan/planner.c
+++ b/src/backend/optimizer/plan/planner.c
@@ -5333,7 +5333,7 @@ plan_cluster_use_sort(Oid tableOid, Oid indexOid)
 	indexScanPath = create_index_path(root, indexInfo,
 									  NIL, NIL, NIL, NIL, NIL,
 									  ForwardScanDirection, false,
-									  NULL, 1.0);
+									  NULL, 1.0, false);
 
 	return (seqScanAndSortPath.total_cost < indexScanPath->path.total_cost);
 }
diff --git a/src/backend/optimizer/util/pathnode.c b/src/backend/optimizer/util/pathnode.c
index f440875ceb..324829690d 100644
--- a/src/backend/optimizer/util/pathnode.c
+++ b/src/backend/optimizer/util/pathnode.c
@@ -744,10 +744,9 @@ add_path_precheck(RelOptInfo *parent_rel,
  *	  As with add_path, we pfree paths that are found to be dominated by
  *	  another partial path; this requires that there be no other references to
  *	  such paths yet.  Hence, GatherPaths must not be created for a rel until
- *	  we're done creating all partial paths for it.  We do not currently build
- *	  partial indexscan paths, so there is no need for an exception for
- *	  IndexPaths here; for safety, we instead Assert that a path to be freed
- *	  isn't an IndexPath.
+ *	  we're done creating all partial paths for it.  Unlike add_path, we don't
+ *	  take an exception for IndexPaths as partial index paths won't be
+ *	  referenced by partial BitmapHeapPaths.
  */
 void
 add_partial_path(RelOptInfo *parent_rel, Path *new_path)
@@ -826,8 +825,6 @@ add_partial_path(RelOptInfo *parent_rel, Path *new_path)
 		{
 			parent_rel->partial_pathlist =
 				list_delete_cell(parent_rel->partial_pathlist, p1, p1_prev);
-			/* we should not see IndexPaths here, so always safe to delete */
-			Assert(!IsA(old_path, IndexPath));
 			pfree(old_path);
 			/* p1_prev does not advance */
 		}
@@ -860,8 +857,6 @@ add_partial_path(RelOptInfo *parent_rel, Path *new_path)
 	}
 	else
 	{
-		/* we should not see IndexPaths here, so always safe to delete */
-		Assert(!IsA(new_path, IndexPath));
 		/* Reject and recycle the new path */
 		pfree(new_path);
 	}
@@ -1005,6 +1000,7 @@ create_samplescan_path(PlannerInfo *root, RelOptInfo *rel, Relids required_outer
  * 'required_outer' is the set of outer relids for a parameterized path.
  * 'loop_count' is the number of repetitions of the indexscan to factor into
  *		estimates of caching behavior.
+ * 'partial_path' is true if constructing a parallel index scan path.
  *
  * Returns the new path node.
  */
@@ -1019,7 +1015,8 @@ create_index_path(PlannerInfo *root,
 				  ScanDirection indexscandir,
 				  bool indexonly,
 				  Relids required_outer,
-				  double loop_count)
+				  double loop_count,
+				  bool partial_path)
 {
 	IndexPath  *pathnode = makeNode(IndexPath);
 	RelOptInfo *rel = index->rel;
@@ -1049,7 +1046,7 @@ create_index_path(PlannerInfo *root,
 	pathnode->indexorderbycols = indexorderbycols;
 	pathnode->indexscandir = indexscandir;
 
-	cost_index(pathnode, root, loop_count);
+	cost_index(pathnode, root, loop_count, partial_path);
 
 	return pathnode;
 }
@@ -3247,7 +3244,7 @@ reparameterize_path(PlannerInfo *root, Path *path,
 				memcpy(newpath, ipath, sizeof(IndexPath));
 				newpath->path.param_info =
 					get_baserel_parampathinfo(root, rel, required_outer);
-				cost_index(newpath, root, loop_count);
+				cost_index(newpath, root, loop_count, false);
 				return (Path *) newpath;
 			}
 		case T_BitmapHeapScan:
diff --git a/src/backend/optimizer/util/plancat.c b/src/backend/optimizer/util/plancat.c
index 7836e6b3f8..4ed27054a1 100644
--- a/src/backend/optimizer/util/plancat.c
+++ b/src/backend/optimizer/util/plancat.c
@@ -241,6 +241,7 @@ get_relation_info(PlannerInfo *root, Oid relationObjectId, bool inhparent,
 			info->amoptionalkey = amroutine->amoptionalkey;
 			info->amsearcharray = amroutine->amsearcharray;
 			info->amsearchnulls = amroutine->amsearchnulls;
+			info->amcanparallel = amroutine->amcanparallel;
 			info->amhasgettuple = (amroutine->amgettuple != NULL);
 			info->amhasgetbitmap = (amroutine->amgetbitmap != NULL);
 			info->amcostestimate = amroutine->amcostestimate;
diff --git a/src/backend/utils/adt/selfuncs.c b/src/backend/utils/adt/selfuncs.c
index fa32e9eabe..d14f0f97a8 100644
--- a/src/backend/utils/adt/selfuncs.c
+++ b/src/backend/utils/adt/selfuncs.c
@@ -6471,7 +6471,8 @@ add_predicate_to_quals(IndexOptInfo *index, List *indexQuals)
 void
 btcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 			   Cost *indexStartupCost, Cost *indexTotalCost,
-			   Selectivity *indexSelectivity, double *indexCorrelation)
+			   Selectivity *indexSelectivity, double *indexCorrelation,
+			   double *indexPages)
 {
 	IndexOptInfo *index = path->indexinfo;
 	List	   *qinfos;
@@ -6761,12 +6762,14 @@ btcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	*indexTotalCost = costs.indexTotalCost;
 	*indexSelectivity = costs.indexSelectivity;
 	*indexCorrelation = costs.indexCorrelation;
+	*indexPages = costs.numIndexPages;
 }
 
 void
 hashcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 				 Cost *indexStartupCost, Cost *indexTotalCost,
-				 Selectivity *indexSelectivity, double *indexCorrelation)
+				 Selectivity *indexSelectivity, double *indexCorrelation,
+				 double *indexPages)
 {
 	List	   *qinfos;
 	GenericCosts costs;
@@ -6807,12 +6810,14 @@ hashcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	*indexTotalCost = costs.indexTotalCost;
 	*indexSelectivity = costs.indexSelectivity;
 	*indexCorrelation = costs.indexCorrelation;
+	*indexPages = costs.numIndexPages;
 }
 
 void
 gistcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 				 Cost *indexStartupCost, Cost *indexTotalCost,
-				 Selectivity *indexSelectivity, double *indexCorrelation)
+				 Selectivity *indexSelectivity, double *indexCorrelation,
+				 double *indexPages)
 {
 	IndexOptInfo *index = path->indexinfo;
 	List	   *qinfos;
@@ -6866,12 +6871,14 @@ gistcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	*indexTotalCost = costs.indexTotalCost;
 	*indexSelectivity = costs.indexSelectivity;
 	*indexCorrelation = costs.indexCorrelation;
+	*indexPages = costs.numIndexPages;
 }
 
 void
 spgcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 				Cost *indexStartupCost, Cost *indexTotalCost,
-				Selectivity *indexSelectivity, double *indexCorrelation)
+				Selectivity *indexSelectivity, double *indexCorrelation,
+				double *indexPages)
 {
 	IndexOptInfo *index = path->indexinfo;
 	List	   *qinfos;
@@ -6925,6 +6932,7 @@ spgcostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	*indexTotalCost = costs.indexTotalCost;
 	*indexSelectivity = costs.indexSelectivity;
 	*indexCorrelation = costs.indexCorrelation;
+	*indexPages = costs.numIndexPages;
 }
 
 
@@ -7222,7 +7230,8 @@ gincost_scalararrayopexpr(PlannerInfo *root,
 void
 gincostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 				Cost *indexStartupCost, Cost *indexTotalCost,
-				Selectivity *indexSelectivity, double *indexCorrelation)
+				Selectivity *indexSelectivity, double *indexCorrelation,
+				double *indexPages)
 {
 	IndexOptInfo *index = path->indexinfo;
 	List	   *indexQuals = path->indexquals;
@@ -7537,6 +7546,7 @@ gincostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	*indexStartupCost += qual_arg_cost;
 	*indexTotalCost += qual_arg_cost;
 	*indexTotalCost += (numTuples * *indexSelectivity) * (cpu_index_tuple_cost + qual_op_cost);
+	*indexPages = dataPagesFetched;
 }
 
 /*
@@ -7545,7 +7555,8 @@ gincostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 void
 brincostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 				 Cost *indexStartupCost, Cost *indexTotalCost,
-				 Selectivity *indexSelectivity, double *indexCorrelation)
+				 Selectivity *indexSelectivity, double *indexCorrelation,
+				 double *indexPages)
 {
 	IndexOptInfo *index = path->indexinfo;
 	List	   *indexQuals = path->indexquals;
@@ -7597,6 +7608,7 @@ brincostestimate(PlannerInfo *root, IndexPath *path, double loop_count,
 	*indexStartupCost += qual_arg_cost;
 	*indexTotalCost += qual_arg_cost;
 	*indexTotalCost += (numTuples * *indexSelectivity) * (cpu_index_tuple_cost + qual_op_cost);
+	*indexPages = index->pages;
 
 	/* XXX what about pages_per_range? */
 }