Simon Riggs <simon@2ndQuadrant.com> writes:
> On Thu, Jan 5, 2012 at 7:26 PM, Robert Haas <robertmhaas@gmail.com> wrote:
>> On the other hand, I think there's a decent argument that he should
>> change his opinion, because 192kB of memory is not a lot. �However,
>> what I mostly want is something that nobody hates, so we can get it
>> committed and move on.
> If that was a reasonable objection it would have applied when we added
> serializable support, or any other SLRU for that matter.
> If memory reduction is a concern to anybody, then a separate patch to
> address *all* issues is required. Blocking this patch makes no sense.
No, your argument is the one that makes no sense. The fact that things
could be made better for low-mem situations is not an argument for
instead making them worse. Which is what going to a fixed value of 32
would do, in return for no benefit that I can see compared to using a
formula of some sort. The details of the formula barely matter, though
I would like to see one that bottoms out at less than 8 buffers so that
there is some advantage gained for low-memory cases.
regards, tom lane