Looks like Robert accidentally answered my question about what version his
results were from off-list. Here's his update:
---
Unfortunately it was 8.2.9, as I realized halfway into the run. Here are
the results from a CVS HEAD checkout last night.
*** Query planning times
q1 (the complex one): 367 ms with default_statistics_target = 10, 379
ms with default_statistics_target = 1000 (3% slowdown)
other 6 queries: 18.48 ms with default_statistics_target = 10, 21.75
ms with default_statistics_target = 1000 (18% slowdown)
comparing to 8.2.9 results:
- q1 was 32% slower with dst = 10, 33% slower with dst = 1000
- other queries were 19% slower with dst = 10, 25% slower with dst = 1000
*** ANALYZE times
10 4.283007
20 6.741417
30 9.157209
40 11.445861
50 13.466374
60 15.402897
70 17.365484
80 19.243842
90 20.905635
100 22.643275
150 29.056517
200 31.878688
250 34.240638
300 36.996304
350 39.921178
400 42.443728
450 44.69833
500 47.553414
550 50.305143
600 52.799597
650 54.961691
700 57.117117
750 58.847345
800 61.60112
850 63.020136
900 65.858213
950 68.072528
1000 69.861321
comparing to 8.2.9 results:
- 3x faster with default_statistics_target = 10
- 4.25x faster with default_statistics_target = 1000
...Robert