Re: Move PinBuffer and UnpinBuffer to atomics

On Fri, Apr 8, 2016 at 10:19 PM, Alexander Korotkov <a.korotkov@postgrespro.ru> wrote:

On Fri, Apr 8, 2016 at 7:39 PM, Andres Freund <andres@anarazel.de> wrote:

As you can see in

http://archives.postgresql.org/message-id/CA%2BTgmoaeRbN%3DZ4oWENLvgGLeHEvGZ_S_Z3KGrdScyKiSvNt3oA%40mail.gmail.com
I'm planning to apply this sometime this weekend, after running some
tests and going over the patch again.

Any chance you could have a look over this?

I took a look at this.  Changes you made look good for me.

I also run test on 4x18 Intel server.

On the top of current master results are following:

clients    TPS

1         12562

2         25604

4         52661

8        103209

10       128599

20       256872

30       365718

40       432749

50       513528

60       684943

70       696050

80       923350

90      1119776

100     1208027

110     1229429

120     1163356

130     1107924

140     1084344

150     1014064

160      961730

170      980743

180      968419

The results are quite discouraging because previously we had about 1.5M TPS in the peak while we have only about 1.2M now.  I found that it's not related to the changes you made in the patch, but it's related to 5364b357 "Increase maximum number of clog buffers".  I'm making same benchmark with 5364b357 reverted.