Dmitry Dolgov <9erthalion6@gmail.com> writes:
>> On Wed, Dec 02, 2020 at 12:58:51PM -0500, Tom Lane wrote:
>> So ... one of the things that's been worrying me about this patch
>> from day one is whether it would create a noticeable performance
>> penalty for existing use-cases. I did a small amount of experimentation
>> about that with the v35 patchset, and it didn't take long at all to
>> find that this:
>> ...
>> is about 15% slower with the patch than with HEAD. I'm not sure
>> what an acceptable penalty might be, but 15% is certainly not it.
> I've tried to reproduce that, but get ~2-4% slowdown (with a pinned
> backend, no turbo etc). Are there any special steps I've probably
> missed?
Hmm, no, I just built with --disable-cassert and otherwise my usual
development options.
I had experimented with some other variants of the test case,
where the repeated statement is
a[i] := i; -- about the same
a[i] := a[i-1] + 1; -- 7% slower
a[i] := a[i-1] - a[i-2]; -- 15% slower
so it seems clear that the penalty is on the array fetch not array
assign side. This isn't too surprising now that I think about it,
because plpgsql's array assignment code is untouched by the patch
(which is a large feature omission BTW: you still can't write
jsonb['x'] := y;
in plpgsql).
regards, tom lane