Re: [HACKERS] Better tracking of free space during SP-GiST index build

2016-09-28 Thread Jeff Janes
On Wed, Sep 28, 2016 at 10:48 AM, Robert Haas wrote: > On Sun, Sep 25, 2016 at 3:28 PM, Tomas Vondra > wrote: > > Sure, that would be useful. > > > > I think it would be useful to make repository of such data sets, so that > > patch authors &

Re: [HACKERS] Better tracking of free space during SP-GiST index build

2016-09-28 Thread Robert Haas
On Wed, Sep 28, 2016 at 2:11 PM, Tom Lane wrote: > Robert Haas writes: >> On Thu, Sep 22, 2016 at 1:37 PM, Tom Lane wrote: >>> OK, I'll think about how to do that more efficiently. The smaller >>> incremental improvement isn't

Re: [HACKERS] Better tracking of free space during SP-GiST index build

2016-09-28 Thread Tom Lane
Robert Haas writes: > On Thu, Sep 22, 2016 at 1:37 PM, Tom Lane wrote: >> OK, I'll think about how to do that more efficiently. The smaller >> incremental improvement isn't surprising, because in this example the >> index would still be 90-something MB

Re: [HACKERS] Better tracking of free space during SP-GiST index build

2016-09-28 Thread Robert Haas
On Thu, Sep 22, 2016 at 1:37 PM, Tom Lane wrote: > OK, I'll think about how to do that more efficiently. The smaller > incremental improvement isn't surprising, because in this example the > index would still be 90-something MB if it had no free space at all, > so there's

Re: [HACKERS] Better tracking of free space during SP-GiST index build

2016-09-28 Thread Robert Haas
On Sun, Sep 25, 2016 at 3:28 PM, Tomas Vondra wrote: > Sure, that would be useful. > > I think it would be useful to make repository of such data sets, so that > patch authors & reviewers can get a reasonable collection of data sets if > needed, instead of scrambling

Re: [HACKERS] Better tracking of free space during SP-GiST index build

2016-09-25 Thread Tomas Vondra
On 09/25/2016 08:33 PM, Oleg Bartunov wrote: On Sat, Sep 24, 2016 at 11:32 PM, Tomas Vondra wrote: On 09/22/2016 07:37 PM, Tom Lane wrote: Tomas Vondra writes: ... I've tried increasing the cache size to 768 entries, with vast

Re: [HACKERS] Better tracking of free space during SP-GiST index build

2016-09-25 Thread Oleg Bartunov
On Sat, Sep 24, 2016 at 11:32 PM, Tomas Vondra wrote: > On 09/22/2016 07:37 PM, Tom Lane wrote: >> >> Tomas Vondra writes: >> >>> ... I've tried increasing the cache size to 768 >>> entries, with vast majority of them (~600) allocated

Re: [HACKERS] Better tracking of free space during SP-GiST index build

2016-09-24 Thread Tomas Vondra
On 09/22/2016 07:37 PM, Tom Lane wrote: Tomas Vondra writes: ... I've tried increasing the cache size to 768 entries, with vast majority of them (~600) allocated to leaf pages. Sadly, this seems to only increase the CREATE INDEX duration a bit, without making the

Re: [HACKERS] Better tracking of free space during SP-GiST index build

2016-09-22 Thread Tom Lane
Tomas Vondra writes: >> On 08/25/2016 01:45 AM, Tom Lane wrote: >>> I'll put this in the commitfest queue. It could use review from >>> someone with the time and motivation to do performance >>> testing/tuning. > I've been toying with this patch a bit today,

Re: [HACKERS] Better tracking of free space during SP-GiST index build

2016-09-21 Thread Tomas Vondra
On 08/25/2016 03:26 AM, Tomas Vondra wrote: On 08/25/2016 01:45 AM, Tom Lane wrote: Over in the thread about the SP-GiST inet opclass, I threatened to post a patch like this, and here it is. The basic idea is to track more than just the very latest page we've used in each of the page

Re: [HACKERS] Better tracking of free space during SP-GiST index build

2016-08-24 Thread Tomas Vondra
On 08/25/2016 01:45 AM, Tom Lane wrote: Over in the thread about the SP-GiST inet opclass, I threatened to post a patch like this, and here it is. The basic idea is to track more than just the very latest page we've used in each of the page categories that SP-GiST works with. I started with

[HACKERS] Better tracking of free space during SP-GiST index build

2016-08-24 Thread Tom Lane
Over in the thread about the SP-GiST inet opclass, I threatened to post a patch like this, and here it is. The basic idea is to track more than just the very latest page we've used in each of the page categories that SP-GiST works with. I started with an arrangement that gave an equal number of