Hi,
David and I were chatting about this patch, in the context of his bump
allocator patch. Attached is a rebased version that is also split up into two
steps, and a bit more polished.
I wasn't sure what a good test was. I ended up measuring
COPY pgbench_accounts TO '/dev/null' WITH (FORMAT 'binary');
of a scale 1 database with pgbench:
c=1;pgbench -q -i -s1 && pgbench -n -c$c -j$c -t100 -f <(echo "COPY pgbench_accounts TO '/dev/null' WITH (FORMAT
'binary');")
average latency
HEAD: 33.865 ms
01: 32.820 ms
02: 29.934 ms
The server was pinned to the one core, turbo mode disabled. That's a pretty
nice win, I'd say. And I don't think this is actually the most allocator
bound workload, I just tried something fairly random...
Greetings,
Andres Freund