andygrove opened a new pull request, #4921:
URL: https://github.com/apache/datafusion-comet/pull/4921
> This PR was created by an LLM as a draft PR. I will mark it as ready for
review after human review.
## Which issue does this PR close?
N/A — autonomous exploratory PR.
## Rationale for this change
Probe the bloom filter with an i64-specialized murmur3, a multiply-based
remainder replacing the per-hash integer division, and a single hoisted
version/modulus dispatch writing straight into a BooleanBuffer instead of
collecting per-row Options.
## What changes are included in this PR?
Probe the bloom filter with an i64-specialized murmur3, a multiply-based
remainder replacing the per-hash integer division, and a single hoisted
version/modulus dispatch writing straight into a BooleanBuffer instead of
collecting per-row Options.
## How are these changes tested?
Correctness: unit tests + seeded differential fuzz (bit-identical Arrow
output vs `main`).
Benchmark (criterion):
- saturated: 17.362% faster (base 122667ns -> cand 101370ns)
- sparse: 23.629% faster (base 78306ns -> cand 59803ns)
Full criterion output:
```text
bloom_filter_might_contain/sparse
time: [59.552 µs 59.748 µs 59.990 µs]
change: [−23.958% −23.629% −23.297%] (p = 0.00 <
0.05)
Performance has improved.
Found 20 outliers among 100 measurements (20.00%)
2 (2.00%) high mild
18 (18.00%) high severe
bloom_filter_might_contain/saturated
time: [101.25 µs 101.28 µs 101.32 µs]
change: [−17.440% −17.362% −17.279%] (p = 0.00 <
0.05)
Performance has improved.
Found 14 outliers among 100 measurements (14.00%)
1 (1.00%) low mild
5 (5.00%) high mild
8 (8.00%) high severe
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
