Re: track_planning causing performance regression

On 2020/07/02 1:54, Andres Freund wrote:
> Hi,
>
> On 2020-07-01 22:20:50 +0900, Fujii Masao wrote:
>> On 2020/07/01 4:03, Andres Freund wrote:
>>> Why did you add the hashing here? It seems a lot better to just add an
>>> lwlock in-place instead of the spinlock? The added size is neglegible
>>> compared to the size of pgssEntry.
>>
>> Because pgssEntry is not array entry but hashtable entry. First I was
>> thinking to assign per-process lwlock to each entry in the array at the
>> startup. But each entry is created every time new entry is required.
>> So lwlock needs to be assigned to each entry at that creation time.
>> We cannnot easily assign lwlock to all the entries at the startup.
>
> But why not just do it exactly at the place the SpinLockInit() is done
> currently?

Sorry I failed to understand your point... You mean that new lwlock should
be initialized at the place the SpinLockInit() is done currently instead of
requesting postmaster to initialize all the lwlocks required for pgss
at _PG_init()?

Regards,

--
Fujii Masao
Advanced Computing Technology Center
Research and Development Headquarters
NTT DATA CORPORATION