RE: [Patch] Optimize dropping of relation buffers using dlist

Hi Amit,

>I think one table with a varying amount of data is sufficient for the vacuum test. 
>I think with more number of tables there is a greater chance of variation. 
>We have previously used multiple tables in one of the tests because of 
>the Truncate operation (which uses DropRelFileNodesAllBuffers that 
>takes multiple relations as input) and that is not true for Vacuum operation which I suppose you are testing here.

I retested performance on single table for several times, the table size is varying with the
BUF_DROP_FULL_SCAN_THRESHOLDfor different shared buffers.
 
When shared_buffers is below 20G, there were no significant changes between master(HEAD) and patched.
And according to the results compared between 20G and 100G, we can get optimization up to NBuffers/128, but there is no
benefitfrom NBuffers/256.
 
I've tested many times, most times the same results came out, I don't know why. But If I used 5 tables(each table size
isset as BUF_DROP_FULL_SCAN_THRESHOLD), then we can get benefit from it(NBuffers/256).
 

Here is my test results for single table. If you have any question or suggestion, kindly let me know.

%reg= (patched- master(HEAD))/ patched
Optimized percentage:
shared_buffers    %reg(NBuffers/512)    %reg(NBuffers/256)    %reg(NBuffers/128)    %reg(NBuffers/64)
%reg(NBuffers/32)   %reg(NBuffers/16)    %reg(NBuffers/8)    %reg(NBuffers/4)
 

----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
128M            0%            0%            -1%            0%            1%            0%            0%            0%
1G            -1%            0%            -1%            0%            0%            0%            0%            0%
20G            0%            0%            -33%            0%            0%            -13%            0%
0%
100G            -32%            0%            -49%            0%            10%            30%            0%
6%

Result details(unit: second):
patched     (sec)                    
shared_buffers    NBuffers/512    NBuffers/256    NBuffers/128    NBuffers/64    NBuffers/32    NBuffers/16
NBuffers/8   NBuffers/4
 
128M        0.107        0.107        0.107        0.107        0.108        0.107        0.108        0.208
1G        0.107        0.107        0.107         0.108         0.208         0.208         0.308         0.409 
20G        0.208         0.308         0.308         0.409         0.609         0.808         1.511         2.713 
100G        0.309         0.408         0.609         1.010         2.011         5.017         6.620         13.931

master(HEAD) (sec)                    
shared_buffers    NBuffers/512    NBuffers/256    NBuffers/128    NBuffers/64    NBuffers/32    NBuffers/16
NBuffers/8   NBuffers/4
 
128M        0.107        0.107        0.108        0.107        0.107        0.107        0.108        0.208
1G        0.108         0.107         0.108         0.108         0.208         0.207         0.308         0.409 
20G        0.208         0.309         0.409         0.409         0.609         0.910         1.511         2.712 
100G        0.408         0.408         0.909         1.010         1.811         3.515         6.619         13.032

Regards
Tang