|From:||Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>|
|To:||Andres Freund <andres(at)anarazel(dot)de>|
|Cc:||Pavel Stehule <pavel(dot)stehule(at)gmail(dot)com>, PostgreSQL Hackers <pgsql-hackers(at)postgresql(dot)org>|
|Subject:||Re: Manipulating complex types as non-contiguous structures in-memory|
|Views:||Raw Message | Whole Thread | Download mbox | Resend email|
Andres Freund <andres(at)anarazel(dot)de> writes:
> On 2015-05-13 21:01:43 -0400, Tom Lane wrote:
>> It is, but why would it be a disaster? We could add StaticAsserts
>> verifying that the sizes actually are different. I doubt that the pad
>> space itself could amount to any issue performance-wise, since it would
>> only ever exist in transient in-memory tuples, and even that only seldom.
> The sizes would be platform dependant.
So what? There are lots of platform-dependent constants in PG.
> It's also just incredibly ugly to
> have to add pad bytes to structures so we can disambiguate them.
Well, I agree it's not too pretty, but you were the one who brought up
the issue of the speed of VARTAG_SIZE(). We definitely gave up some
performance there already, and my patch will make it worse.
> Anyway, I think we can live with your & or my proposed additional branch
> for now. I can't see either variant being a relevant performance
> bottleneck anytime soon.
Actually, after having microbenchmarked the difference between those
two proposals, I'm not too sure that VARTAG_SIZE() is down in the noise.
But it doesn't matter for the moment --- any one of these alternatives
would be a very localized code change, and none of them would create
an on-disk compatibility break. We can let it go until someone wants
to put together a more definitive benchmark for testing.
regards, tom lane
|Next Message||Robert Haas||2015-05-14 01:39:03||upper planner path-ification|
|Previous Message||Andres Freund||2015-05-14 01:06:09||Re: Manipulating complex types as non-contiguous structures in-memory|