pgrust v0.2 is not fast when sum a series

[l1t1]

pgrust v0.2

docker run -d --name pgrust02 -e POSTGRES_PASSWORD=secret -p 5432:5432 malisper/pgrust:v0.2
90e1170ad8e5eed1b04bd297c27ebc7aef62cf6e634d671c98efb853ec33b05f
psql -h localhost -p 5432 -U postgres
postgres=# \timing on
Timing is on.
postgres=# select sum(i) from generate_series(1,100000000)t(i);
       sum
------------------
 5000000050000000
(1 row)

Time: 11397.259 ms (00:11.397)

postgresql 17.7

docker start pg177
pg177
root@DESKTOP-59T6U68:/mnt/c/d# docker exec -ti pg177 bash
root@b3490d8427cd:/# su - postgres
postgres@b3490d8427cd:~$ psql
psql (17.7 (Debian 17.7-3.pgdg13+1))
Type "help" for help.

postgres=# \timing on
Timing is on.
postgres=# select sum(i) from generate_series(1,100000000)t(i);
       sum
------------------
 5000000050000000
(1 row)

Time: 8651.461 ms (00:08.651)

duckdb

./duckdb155
DuckDB v1.5.5 (Variegata)
Enter ".help" for usage hints.
memory D .timer on
memory D select sum(i) from generate_series(1,100000000)t(i);
┌────────────────────┐
│       sum(i)       │
│       int128       │
├────────────────────┤
│  5000000050000000  │
│ (5.00 quadrillion) │
└────────────────────┘
Run Time (s): real 0.103 user 0.173758 sys 0.000000

[l1t1]

clickhouse

./clickhouse
Decompressing the binary......
ClickHouse local version 26.8.1.1060 (official build).
:) select sum(generate_series) from generate_series(1,100000000) a;

SELECT sum(generate_series)
FROM generate_series(1, 100000000) AS a

Query id: 9849b72e-692d-446b-94af-5cde915704ec

   ┌─sum(generate_series)─┐
1. │     5000000050000000 │ -- 5.00 quadrillion
   └──────────────────────┘

1 row in set. Elapsed: 0.099 sec. Processed 100.00 million rows, 800.00 MB (1.01 billion rows/s., 8.11 GB/s.)
Peak memory usage: 52.28 KiB.

[malisper]

Re-measured on current main (fleet, Graviton, PostgreSQL 18.4 client/server on the same data directory), dist profile:

statement	C 18 (s)	pgrust (s)	ratio
SELECT sum(g) FROM generate_series(1, 50000000) g	5.45	5.14	0.94
SELECT count(*) FROM generate_series(1, 50000000)	5.08	3.86	0.76
SELECT sum(g) FROM (SELECT g FROM generate_series(1, 20000000) g ORDER BY g DESC) x	5.06	4.96	0.98

Results are byte-identical (fleet job pgrust-fast-tests-070ae96afe-1789010874-6fd3).

One more thing this surfaced: memory. The v0.2-era run of the first statement held ~1.5 GB RSS (a per-row copy of every spilled tuplestore row into the query arena). That is fixed in Pager-Free/pgrust-fast#2062 (tuplestore READFILE reads lend the staged image to the slot instead of copying); with it the same statements grow RSS by 10-22 MB, within the work_mem-bounded tuplestore like C (pgrust-fast-tests-311b09c185-1789011797-36bf).

Regression guard: scripts/generate-series-perf-e2e.sh (byte-identity, a ≤2.0x wall-time ratio gate on dist builds, and a ≤256 MB RSS-growth gate), landing with that PR. Closing.