On Fri, Nov 18, 2022 at 1:50 PM Peter J. Holzer wrote:
> There should be about 27000 of them, same as for the othe index, right?
There aren't that many. The point I'm making is that you can access
each VM page approximately once (and check relatively many index
tuple's TIDs all in one go), or man
On 2022-11-18 13:09:16 -0800, Peter Geoghegan wrote:
> On Fri, Nov 18, 2022 at 12:46 PM Peter J. Holzer wrote:
> > Both do a parallel index only scan. Both perform 0 heap fetches.
> > But one reads 27336 buffers (or about 22 bytes per index entry, which
> > sounds reasonable) while the other reads
On 2022-11-18 15:59:46 -0500, Tom Lane wrote:
> "Peter J. Holzer" writes:
> > Both do a parallel index only scan. Both perform 0 heap fetches.
> > But one reads 27336 buffers (or about 22 bytes per index entry, which
> > sounds reasonable) while the other reads 9995216 buffers (or almost one
> > f
On Fri, Nov 18, 2022 at 12:46 PM Peter J. Holzer wrote:
> Both do a parallel index only scan. Both perform 0 heap fetches.
> But one reads 27336 buffers (or about 22 bytes per index entry, which
> sounds reasonable) while the other reads 9995216 buffers (or almost one
> full buffer per row). Why?
"Peter J. Holzer" writes:
> Both do a parallel index only scan. Both perform 0 heap fetches.
> But one reads 27336 buffers (or about 22 bytes per index entry, which
> sounds reasonable) while the other reads 9995216 buffers (or almost one
> full buffer per row). Why? The entries should be dense in
