On 17 November 2014 05:49, Rajeev rastogi <rajeev.rastogi@huawei.com> wrote:
>> * Single column bigint index
> It gives around 5% improvement.
I confirm 5% improvement in both short and long index scans, on the
least beneficial datatype. Looks likely to be a very positive win
overall.
TEST 1: 100000x
select id2 into x from tbl2 where id2>990000 limit 1;
Patched
Time: 617.026 ms
Time: 604.769 ms
Time: 613.392 ms
Time: 600.484 ms
Time: 607.895 ms
Time: 609.366 ms
Time: 606.187 ms
Time: 602.643 ms
Time: 613.115 ms
Time: 605.705 ms
Unpatched
Time: 642.786 ms
Time: 649.880 ms
Time: 637.400 ms
Time: 635.937 ms
Time: 641.210 ms
Time: 641.988 ms
Time: 637.839 ms
Time: 640.265 ms
Time: 640.237 ms
Time: 637.669 ms
TEST 2 10000x
select count(*) into x from tbl2 where id2>990000;
Patched:
Time: 17277.083 ms
Time: 17373.752 ms
Time: 17208.220 ms
Unpatched:
Time: 18186.062 ms
Time: 18067.664 ms
Time: 18118.807 ms
>> The explanatory comments need some work to more clearly explain what
>> this patch does.
>
> Please help me to understand this point, you want me to add more comments about patch in this mail chain or in code.
The language used in the comments is not clear enough. I'll do my best
to improve that, then look to commit this in about 5 hours.
-- Simon Riggs http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services