Github user saketj commented on the issue:
https://github.com/apache/incubator-quickstep/pull/114
It turns out that these queries have very few scan (selection) operations
as compared to other queries and mostly involve joins where BitVectors might
not be aggressively used. Also, if the selectivity is high, branch predictions
will anyway be more-or-less accurate. (more so after the LIP filter
optimization).
However, another interesting thing is that for these queries 09, 20, 21,
the differences are actually not high across different trials. A cold cache may
be playing some tricks, when we report our aggregate numbers.
| Query 9 | w/o PR (ms) | w/ PR (ms) |
|----------|-------------|------------|
| Trial 1 | 110251 | 42551 |
| Trial 2 | 13763 | 14386 |
| Trial 3 | 10302 | 10129 |
| Trial 4 | 11783 | 10616 |
| Trial 5 | 10070 | 9243 |
| | | |
| | | |
| Query 20 | w/o PR (ms) | w/ PR (ms) |
| Trial 1 | 85988 | 104532 |
| Trial 2 | 56639 | 58389 |
| Trial 3 | 50041 | 51712 |
| Trial 4 | 54269 | 50740 |
| Trial 5 | 49688 | 50723 |
| | | |
| | | |
| Query 21 | w/o PR (ms) | w/ PR (ms) |
| Trial 1 | 126323 | 142742 |
| Trial 2 | 120334 | 119632 |
| Trial 3 | 126304 | 115668 |
| Trial 4 | 115185 | 114618 |
| Trial 5 | 114663 | 113232 |
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
