Introduce CompactAttribute array in TupleDesc

The new compact_attrs array stores a few select fields from
FormData_pg_attribute in a more compact way, using only 16 bytes per
column instead of the 104 bytes that FormData_pg_attribute uses.  Using
CompactAttribute allows performance-critical operations such as tuple
deformation to be performed without looking at the FormData_pg_attribute
element in TupleDesc which means fewer cacheline accesses.  With this
change, NAMEDATALEN could be increased with a much smaller negative impact
on performance.

For some workloads, tuple deformation can be the most CPU intensive part
of processing the query.  Some testing with 16 columns on a table
where the first column is variable length showed around a 10% increase in
transactions per second for an OLAP type query performing aggregation on
the 16th column.  However, in certain cases, the increases were much
higher, up to ~25% on one AMD Zen4 machine.

This also makes pg_attribute.attcacheoff redundant.  A follow-on commit
will remove it, thus shrinking the FormData_pg_attribute struct by 4
bytes.

Author: David Rowley
Discussion: https://postgr.es/m/CAApHDvrBztXP3yx=NKNmo3xwFAFhEdyPnvrDg3=M0RhDs+4vYw@mail.gmail.com
Reviewed-by: Andres Freund, Victor Yegorov

1 files changed, 13 insertions, 0 deletions

diff --git a/src/backend/utils/cache/typcache.c b/src/backend/utils/cache/typcache.c
index 6590cbe5c67..8ce32a475f0 100644
--- a/src/backend/utils/cache/typcache.c
+++ b/src/backend/utils/cache/typcache.c
@@ -241,12 +241,18 @@ shared_record_table_compare(const void *a, const void *b, size_t size,
 	TupleDesc	t2;
 
 	if (k1->shared)
+	{
 		t1 = (TupleDesc) dsa_get_address(area, k1->u.shared_tupdesc);
+		t1->attrs = TupleDescAttrAddress(t1);
+	}
 	else
 		t1 = k1->u.local_tupdesc;
 
 	if (k2->shared)
+	{
 		t2 = (TupleDesc) dsa_get_address(area, k2->u.shared_tupdesc);
+		t2->attrs = TupleDescAttrAddress(t2);
+	}
 	else
 		t2 = k2->u.local_tupdesc;
 
@@ -264,7 +270,10 @@ shared_record_table_hash(const void *a, size_t size, void *arg)
 	TupleDesc	t;
 
 	if (k->shared)
+	{
 		t = (TupleDesc) dsa_get_address(area, k->u.shared_tupdesc);
+		t->attrs = TupleDescAttrAddress(t);
+	}
 	else
 		t = k->u.local_tupdesc;
 
@@ -1867,6 +1876,7 @@ lookup_rowtype_tupdesc_internal(Oid type_id, int32 typmod, bool noError)
 					tupdesc = (TupleDesc)
 						dsa_get_address(CurrentSession->area,
 										entry->shared_tupdesc);
+					tupdesc->attrs = TupleDescAttrAddress(tupdesc);
 					Assert(typmod == tupdesc->tdtypmod);
 
 					/* We may need to extend the local RecordCacheArray. */
@@ -2953,6 +2963,7 @@ find_or_make_matching_shared_tupledesc(TupleDesc tupdesc)
 		result = (TupleDesc)
 			dsa_get_address(CurrentSession->area,
 							record_table_entry->key.u.shared_tupdesc);
+		result->attrs = TupleDescAttrAddress(result);
 		Assert(result->tdrefcount == -1);
 
 		return result;
@@ -3016,6 +3027,7 @@ find_or_make_matching_shared_tupledesc(TupleDesc tupdesc)
 		result = (TupleDesc)
 			dsa_get_address(CurrentSession->area,
 							record_table_entry->key.u.shared_tupdesc);
+		result->attrs = TupleDescAttrAddress(result);
 		Assert(result->tdrefcount == -1);
 
 		return result;
@@ -3028,6 +3040,7 @@ find_or_make_matching_shared_tupledesc(TupleDesc tupdesc)
 						record_table_entry);
 	result = (TupleDesc)
 		dsa_get_address(CurrentSession->area, shared_dp);
+	result->attrs = TupleDescAttrAddress(result);
 	Assert(result->tdrefcount == -1);
 
 	return result;