FAQ

Update of /usr/local/cvsroot/postgres95/src/backend/tcop
In directory hub.org:/home/users/v/vadim/postgres95/src/backend/tcop

Modified Files:
variable.c
Log Message:
New VAR r_plans added to enable turn ON/OFF
using right-sided plans.
Vadim, are we keeping this. Do we need to add it to the manual pages
and psql help?

What are we doing with GEQO and when it gets invoked. Do we need to
update the manual page to explain it?

- --
Bruce Momjian
maillist@candle.pha.pa.us

------------------------------

Search Discussions

  • Vadim B. Mikheev at May 23, 1997 at 2:09 am

    Bruce Momjian wrote:
    Update of /usr/local/cvsroot/postgres95/src/backend/tcop
    In directory hub.org:/home/users/v/vadim/postgres95/src/backend/tcop

    Modified Files:
    variable.c
    Log Message:
    New VAR r_plans added to enable turn ON/OFF
    using right-sided plans.
    Vadim, are we keeping this. Do we need to add it to the manual pages
    and psql help?
    I hope that r_plans is temporary option - it will be unuseful
    when we'll got bushy plans... But for the moment we keep it
    and it's good to explain it.
    Note, if defined USE_RIGHT_SIDED_PLANS then they will be used
    by default.
    What are we doing with GEQO and when it gets invoked. Do we need to
    update the manual page to explain it?
    My suggestion is SET geqo TO 'on[=XXX]'. If =XXX omitted then
    use default == 6 (i.e. >= 6 relations), as pointed by Martin.
    If GEQO turned OFF (_use_geqo_ is FALSE) then use old optimizer,
    unconditionally.
    Note, if GEQO defined then _use_geqo_ is TRUE by default.

    Also, Keith wrote:
    I also think MIN_POOL in "include/optimizer/geqo.h:46" could do with
    reducing so that the computed minimum can be more realistic when there
    are a small number of relations.( I reduced it to 6 and pool was computed
    as 8 for a 2 rel join, 16 for a 3 rel join)

    If there is no pg_geqo config file the computed value of effort in
    "backend/optimizer/geqo/geqo_params.c:215" seems to be very rough.
    ie. if PoolSize is MAX_POOL (1024) we use HIGH effort otherwize we
    use MEDIUM effort. On my small joins LOW effort gives the best results.
    I can't comment on this. Martin ?

    Vadim

    ------------------------------
  • Martin S. Utesch at May 23, 1997 at 8:32 am

    Vadim wrote:


    Also, Keith wrote:
    I also think MIN_POOL in "include/optimizer/geqo.h:46" could do with
    reducing so that the computed minimum can be more realistic when there
    are a small number of relations.( I reduced it to 6 and pool was computed
    as 8 for a 2 rel join, 16 for a 3 rel join)

    If there is no pg_geqo config file the computed value of effort in
    "backend/optimizer/geqo/geqo_params.c:215" seems to be very rough.
    ie. if PoolSize is MAX_POOL (1024) we use HIGH effort otherwize we
    use MEDIUM effort. On my small joins LOW effort gives the best results.
    I can't comment on this. Martin ?

    Vadim
    I played with the parameter settings lots of time, read articels about
    population sizing for GAs.
    The present parameter settings seem sufficiently accurate to me.
    I don't expect changes to have much impact on the performance.

    What is the time payoff with LOW effort for small join queries?

    Recently I proposed to use the standard optimizer approach for small
    queries anyway ...

    =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
    * Martin Utesch * Institute of Automatic Control *
    = = University of Mining and Technology =
    * utesch@aut.tu-freiberg.de * Freiberg, Germany *
    =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=

    ------------------------------

Related Discussions

Discussion Navigation
viewthread | post
Discussion Overview
grouppgsql-hackers @
categoriespostgresql
postedMay 22, '97 at 6:44p
activeMay 23, '97 at 8:32a
posts3
users3
websitepostgresql.org...
irc#postgresql

People

Translate

site design / logo © 2022 Grokbase