Horton

RE: AIX support

Srirama Kucherlapati <sriram.rk@in.ibm.com>

View thread
From: Srirama Kucherlapati <sriram.rk@in.ibm.com>
To: Heikki Linnakangas <hlinnaka@iki.fi>, Heikki Linnakangas <hlinnaka@gmail.com>, Bruce Momjian <bruce@momjian.us>
Cc: Peter Eisentraut <peter@eisentraut.org>, Alvaro Herrera <alvherre@alvh.no-ip.org>, "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>, Laurenz Albe <laurenz.albe@cybertec.at>, Noah Misch <noah@leadboat.com>, Michael Paquier <michael@paquier.xyz>, Andres Freund <andres@anarazel.de>, Thomas Munro <thomas.munro@gmail.com>, "tvk1271@gmail.com" <tvk1271@gmail.com>, "postgres-ibm-aix@wwpdl.vnet.ibm.com" <postgres-ibm-aix@wwpdl.vnet.ibm.com>, Tom Lane <tgl@sss.pgh.pa.us>
Date: 2024-10-04T15:04:57Z
Lists: pgsql-hackers 
Hi Heikki and team,

A few updates…

  > > Ok, if we don't need the assembler code at all, that's good. A patch to
  > > introduce AIX support should not change it for non-AIX powerpc systems
  > > though. That might be a good change, but would need to be justified
  > > separately, e.g.  by some performance testing, and should be a separate
  > > patch.

We ran pgbench for both the patches on powerpc Linux and below are the test
results and both results are close.

PPCLE sync: is the patch with the spin lock using gcc __sync* routines.
PPCLE asm: is the patch with the assemble code.

>> pgbench -c 100 -p 5432 -d postgres -T 180 -r -P 10  -L 10 -j 20


    OS/type             : PPCLE sync        PPCLE asm
    ---------------------------------------------------------
    latency average     : 136.257           138.552
    ---------------------------------------------------------
    latency stddev      : 234.74            238.603
    ---------------------------------------------------------
    Initial
    connection time     : 101.791           88.411
    ---------------------------------------------------------
    TPS(without initial
     connection time)   : 733.633924        721.440648
    ---------------------------------------------------------
    No. of transactions
    actually processed  : 132080            129893
    ---------------------------------------------------------
    No of transactions
    above the 10.0 ms
    latency limit       : 124235/132080     122183/129893
                          (94.060%)         (94.064%)
    ---------------------------------------------------------


Please let us know your comments.
BTW we are working on the other review comments as well.


Warm regards,
Sriram.