Thread: next XID is in shmem now...

next XID is in shmem now...

From
Vadim Mikheev
Date:
Backends fetch 1024 XIDs now and place them in shmem.
There is space in VariableCache struct for OIDs as well
but I didn't change GetNewObjectId() due to the
CheckMaxObjectId() stuff... Bruce ?

All other LLL stuff will be #ifdef-ed...

Vadim

Re: [HACKERS] next XID is in shmem now...

From
Bruce Momjian
Date:
> Backends fetch 1024 XIDs now and place them in shmem.
> There is space in VariableCache struct for OIDs as well
> but I didn't change GetNewObjectId() due to the
> CheckMaxObjectId() stuff... Bruce ?

What can I do to help?  Is the problem that a backend can set the next
oid by specifiying an oid greater than the current one?

>
> All other LLL stuff will be #ifdef-ed...

As far as I am concerned, you don't need use #ifdef.

--
Bruce Momjian                          |  830 Blythe Avenue
maillist@candle.pha.pa.us              |  Drexel Hill, Pennsylvania 19026
  +  If your life is a hard drive,     |  (610) 353-9879(w)
  +  Christ can be your backup.        |  (610) 853-3000(h)

Re: [HACKERS] next XID is in shmem now...

From
Vadim Mikheev
Date:
Bruce Momjian wrote:
>
> > Backends fetch 1024 XIDs now and place them in shmem.
> > There is space in VariableCache struct for OIDs as well
> > but I didn't change GetNewObjectId() due to the
> > CheckMaxObjectId() stuff... Bruce ?
>
> What can I do to help?  Is the problem that a backend can set the next
> oid by specifiying an oid greater than the current one?

No problem - I just havn't time to think about this, sorry.

>
> >
> > All other LLL stuff will be #ifdef-ed...
>
> As far as I am concerned, you don't need use #ifdef.

I'm not sure how much ready/robust this will be in 6.4.
This is long-term project...

Vadim

Re: [HACKERS] next XID is in shmem now...

From
Bruce Momjian
Date:
> Bruce Momjian wrote:
> >
> > > Backends fetch 1024 XIDs now and place them in shmem.
> > > There is space in VariableCache struct for OIDs as well
> > > but I didn't change GetNewObjectId() due to the
> > > CheckMaxObjectId() stuff... Bruce ?
> >
> > What can I do to help?  Is the problem that a backend can set the next
> > oid by specifiying an oid greater than the current one?
>
> No problem - I just havn't time to think about this, sorry.
>
> >
> > >
> > > All other LLL stuff will be #ifdef-ed...
> >
> > As far as I am concerned, you don't need use #ifdef.
>
> I'm not sure how much ready/robust this will be in 6.4.
> This is long-term project...

Any chance on getting the 30-second pg_log syncing, so we can improve
the default pgsql performance, and not do fsync on every transaction by
default?


--
Bruce Momjian                          |  830 Blythe Avenue
maillist@candle.pha.pa.us              |  Drexel Hill, Pennsylvania 19026
  +  If your life is a hard drive,     |  (610) 353-9879(w)
  +  Christ can be your backup.        |  (610) 853-3000(h)