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

src/include/access/htup_details.h

@@ -758,9 +758,9 @@ fastgetattr(HeapTuple tup, int attnum, TupleDesc tupleDesc, bool *isnull)
 	*isnull = false;
 	if (HeapTupleNoNulls(tup))
 	{
-		Form_pg_attribute att;
+		CompactAttribute *att;
 
-		att = TupleDescAttr(tupleDesc, attnum - 1);
+		att = TupleDescCompactAttr(tupleDesc, attnum - 1);
 		if (att->attcacheoff >= 0)
 			return fetchatt(att, (char *) tup->t_data + tup->t_data->t_hoff +
 							att->attcacheoff);

src/include/access/itup.h

@@ -124,11 +124,13 @@ index_getattr(IndexTuple tup, int attnum, TupleDesc tupleDesc, bool *isnull)
 
 	if (!IndexTupleHasNulls(tup))
 	{
-		if (TupleDescAttr(tupleDesc, attnum - 1)->attcacheoff >= 0)
+		CompactAttribute *attr = TupleDescCompactAttr(tupleDesc, attnum - 1);
+
+		if (attr->attcacheoff >= 0)
 		{
-			return fetchatt(TupleDescAttr(tupleDesc, attnum - 1),
-							(char *) tup + IndexInfoFindDataOffset(tup->t_info)
-							+ TupleDescAttr(tupleDesc, attnum - 1)->attcacheoff);
+			return fetchatt(attr,
+							(char *) tup + IndexInfoFindDataOffset(tup->t_info) +
+							attr->attcacheoff);
 		}
 		else
 			return nocache_index_getattr(tup, attnum, tupleDesc);

src/include/access/tupdesc.h

@@ -45,6 +45,39 @@ typedef struct TupleConstr
 	bool		has_generated_stored;
 } TupleConstr;
 
+/*
+ * CompactAttribute
+ *		Cut-down version of FormData_pg_attribute for faster access for tasks
+ *		such as tuple deformation.  These values are populated using the
+ *		populate_compact_attribute function, which must be called directly
+ *		after the FormData_pg_attribute struct is populated or altered in any
+ *		way.
+ *
+ * Currently, this struct is 16 bytes.  Any code changes which enlarge this
+ * struct should be considered very carefully.
+ *
+ * Code which must access a TupleDesc's attribute data should always make use
+ * of the CompactAttribute when the required fields are available there.  It's
+ * more efficient to access the memory in CompactAttribute due to it both
+ * being a more compact representation of FormData_pg_attribute, but also
+ * because accessing the FormData_pg_attribute requires an additional pointer
+ * indirection through TupleDescData.attrs
+ */
+typedef struct CompactAttribute
+{
+	int32		attcacheoff;	/* fixed offset into tuple, if known, or -1 */
+	int16		attlen;			/* attr len in bytes or -1 = varlen, -2 =
+								 * cstring */
+	bool		attbyval;		/* as FormData_pg_attribute.attbyval */
+	bool		attispackable;	/* FormData_pg_attribute.attstorage !=
+								 * TYPSTORAGE_PLAIN */
+	bool		atthasmissing;	/* as FormData_pg_attribute.atthasmissing */
+	bool		attisdropped;	/* as FormData_pg_attribute.attisdropped */
+	bool		attgenerated;	/* FormData_pg_attribute.attgenerated != '\0' */
+	bool		attnotnull;		/* as FormData_pg_attribute.attnotnull */
+	char		attalign;		/* alignment requirement */
+} CompactAttribute;
+
 /*
  * This struct is passed around within the backend to describe the structure
  * of tuples.  For tuples coming from on-disk relations, the information is
@@ -75,6 +108,18 @@ typedef struct TupleConstr
  * context and go away when the context is freed.  We set the tdrefcount
  * field of such a descriptor to -1, while reference-counted descriptors
  * always have tdrefcount >= 0.
+ *
+ * The attrs field stores the fixed-sized portion of FormData_pg_attribute.
+ * Because that struct is large, we also store a corresponding
+ * CompactAttribute for each attribute in compact_attrs.  compact_attrs is
+ * stored inline with the struct.  Because CompactAttribute is significantly
+ * smaller than FormData_pg_attribute, code, especially performance-critical
+ * code, should prioritize using the fields from the CompactAttribute over the
+ * equivalent fields in FormData_pg_attribute whenever possible.
+ *
+ * Any code making changes manually to the fields in 'attrs' must subsequently
+ * call populate_compact_attribute() to flush the changes out to the
+ * corresponding 'compact_attrs' element.
  */
 typedef struct TupleDescData
 {
@@ -84,13 +129,53 @@ typedef struct TupleDescData
 	int			tdrefcount;		/* reference count, or -1 if not counting */
 	TupleConstr *constr;		/* constraints, or NULL if none */
 	/* attrs[N] is the description of Attribute Number N+1 */
-	FormData_pg_attribute attrs[FLEXIBLE_ARRAY_MEMBER];
+	FormData_pg_attribute *attrs;
+	CompactAttribute compact_attrs[FLEXIBLE_ARRAY_MEMBER];
 }			TupleDescData;
 typedef struct TupleDescData *TupleDesc;
 
-/* Accessor for the i'th attribute of tupdesc. */
+extern void populate_compact_attribute(TupleDesc tupdesc, int attnum);
+
+/* Accessor for the i'th FormData_pg_attribute of tupdesc. */
 #define TupleDescAttr(tupdesc, i) (&(tupdesc)->attrs[(i)])
 
+/*
+ * Accessor for the i'th CompactAttribute of tupdesc.
+ */
+static inline CompactAttribute *
+TupleDescCompactAttr(TupleDesc tupdesc, int i)
+{
+	CompactAttribute *cattr = &tupdesc->compact_attrs[i];
+#ifdef USE_ASSERT_CHECKING
+	CompactAttribute snapshot;
+
+	/*
+	 * In Assert enabled builds we verify that the CompactAttribute is
+	 * populated correctly.  This helps find bugs in places such as ALTER
+	 * TABLE where code makes changes to the FormData_pg_attribute but forgets
+	 * to call populate_compact_attribute.
+	 */
+
+	/*
+	 * Take a snapshot of how the CompactAttribute is now before calling
+	 * populate_compact_attribute to make it up-to-date with the
+	 * FormData_pg_attribute.
+	 */
+	memcpy(&snapshot, cattr, sizeof(CompactAttribute));
+
+	populate_compact_attribute(tupdesc, i);
+
+	/* reset attcacheoff back to what it was */
+	cattr->attcacheoff = snapshot.attcacheoff;
+
+	/* Ensure the snapshot matches the freshly populated CompactAttribute */
+	Assert(memcmp(&snapshot, cattr, sizeof(CompactAttribute)) == 0);
+#endif
+
+	return cattr;
+}
+
+
 extern TupleDesc CreateTemplateTupleDesc(int natts);
 
 extern TupleDesc CreateTupleDesc(int natts, Form_pg_attribute *attrs);
@@ -100,9 +185,15 @@ extern TupleDesc CreateTupleDescCopy(TupleDesc tupdesc);
 extern TupleDesc CreateTupleDescCopyConstr(TupleDesc tupdesc);
 
 #define TupleDescSize(src) \
-	(offsetof(struct TupleDescData, attrs) + \
+	(offsetof(struct TupleDescData, compact_attrs) + \
+	 (src)->natts * sizeof(CompactAttribute) + \
 	 (src)->natts * sizeof(FormData_pg_attribute))
 
+#define TupleDescAttrAddress(desc) \
+	(Form_pg_attribute) ((char *) (desc) + \
+	 (offsetof(struct TupleDescData, compact_attrs) + \
+	 (desc)->natts * sizeof(CompactAttribute)))
+
 extern void TupleDescCopy(TupleDesc dst, TupleDesc src);
 
 extern void TupleDescCopyEntry(TupleDesc dst, AttrNumber dstAttno,

src/include/access/tupmacs.h

@@ -14,6 +14,7 @@
 #ifndef TUPMACS_H
 #define TUPMACS_H
 
+#include "access/tupdesc.h"
 #include "catalog/pg_type_d.h"	/* for TYPALIGN macros */
 
 
@@ -30,8 +31,8 @@ att_isnull(int ATT, const bits8 *BITS)
 
 #ifndef FRONTEND
 /*
- * Given a Form_pg_attribute and a pointer into a tuple's data area,
- * return the correct value or pointer.
+ * Given a Form_pg_attribute or CompactAttribute and a pointer into a tuple's
+ * data area, return the correct value or pointer.
  *
  * We return a Datum value in all cases.  If the attribute has "byval" false,
  * we return the same pointer into the tuple data area that we're passed.