Consider parallel merge joins.

Commit 45be99f8 took the position
that performing a merge join in parallel was not likely to work out
well, but this conclusion was greeted with skepticism even at the
time.  Whether it was true then or not, it's clearly not true any
more now that we have parallel index scan.

Dilip Kumar, reviewed by Amit Kapila and by me.

Discussion: http://postgr.es/m/CAFiTN-v3=cM6nyFwFGp0fmvY4=kk79Hq9Fgu0u8CSJ-EEq1Tiw@mail.gmail.com

src/test/regress/expected/select_parallel.out

@@ -169,6 +169,31 @@ select  count(*) from tenk1 where thousand > 95;
 
 reset enable_seqscan;
 reset enable_bitmapscan;
+-- test parallel merge join path.
+set enable_hashjoin to off;
+set enable_nestloop to off;
+explain (costs off)
+	select  count(*) from tenk1, tenk2 where tenk1.unique1 = tenk2.unique1;
+                                  QUERY PLAN                                   
+-------------------------------------------------------------------------------
+ Finalize Aggregate
+   ->  Gather
+         Workers Planned: 4
+         ->  Partial Aggregate
+               ->  Merge Join
+                     Merge Cond: (tenk1.unique1 = tenk2.unique1)
+                     ->  Parallel Index Only Scan using tenk1_unique1 on tenk1
+                     ->  Index Only Scan using tenk2_unique1 on tenk2
+(8 rows)
+
+select  count(*) from tenk1, tenk2 where tenk1.unique1 = tenk2.unique1;
+ count 
+-------
+ 10000
+(1 row)
+
+reset enable_hashjoin;
+reset enable_nestloop;
 set force_parallel_mode=1;
 explain (costs off)
   select stringu1::int2 from tenk1 where unique1 = 1;

src/test/regress/sql/select_parallel.sql

@@ -64,6 +64,16 @@ select  count(*) from tenk1 where thousand > 95;
 reset enable_seqscan;
 reset enable_bitmapscan;
 
+-- test parallel merge join path.
+set enable_hashjoin to off;
+set enable_nestloop to off;
+
+explain (costs off)
+	select  count(*) from tenk1, tenk2 where tenk1.unique1 = tenk2.unique1;
+select  count(*) from tenk1, tenk2 where tenk1.unique1 = tenk2.unique1;
+
+reset enable_hashjoin;
+reset enable_nestloop;
 set force_parallel_mode=1;
 
 explain (costs off)