Jefffrey commented on PR #20044:
URL: https://github.com/apache/datafusion/pull/20044#issuecomment-3821740567
> Thanks @Jefffrey great PR and appreciate commented changes, WDYT about hex
bench results? `datafusion/spark/benches/hex.rs`
I ran the bench on my M4 Mac:
<details>
```
Running benches/hex.rs
(/Users/jeffrey/.cargo_target_cache/release/deps/hex-093ddc801ba8123c)
Gnuplot not found, using plotters backend
hex_int64/size=1024 time: [6.3674 µs 6.4616 µs 6.5677 µs]
change: [+5.4026% +8.6451% +12.018%] (p = 0.00 <
0.05)
Performance has regressed.
hex_int64/size=4096 time: [27.309 µs 27.679 µs 28.046 µs]
change: [−1.2305% +0.6185% +2.2777%] (p = 0.49 >
0.05)
No change in performance detected.
Found 2 outliers among 100 measurements (2.00%)
2 (2.00%) high mild
hex_int64/size=8192 time: [57.803 µs 58.466 µs 59.090 µs]
change: [+5.0492% +5.9280% +6.9853%] (p = 0.00 <
0.05)
Performance has regressed.
hex_utf8/size=1024 time: [41.243 µs 41.363 µs 41.477 µs]
change: [−0.1689% +0.1669% +0.4908%] (p = 0.32 >
0.05)
No change in performance detected.
Found 13 outliers among 100 measurements (13.00%)
1 (1.00%) low severe
12 (12.00%) high mild
hex_utf8/size=4096 time: [163.62 µs 164.08 µs 164.50 µs]
change: [−0.4326% −0.0739% +0.2639%] (p = 0.68 >
0.05)
No change in performance detected.
Found 10 outliers among 100 measurements (10.00%)
3 (3.00%) low mild
7 (7.00%) high mild
hex_utf8/size=8192 time: [326.48 µs 327.31 µs 328.12 µs]
change: [−0.4520% −0.1235% +0.1939%] (p = 0.45 >
0.05)
No change in performance detected.
Found 4 outliers among 100 measurements (4.00%)
1 (1.00%) low mild
3 (3.00%) high mild
hex_binary/size=1024 time: [41.158 µs 41.225 µs 41.286 µs]
change: [−0.4329% −0.1577% +0.1180%] (p = 0.27 >
0.05)
No change in performance detected.
Found 7 outliers among 100 measurements (7.00%)
1 (1.00%) low severe
3 (3.00%) low mild
1 (1.00%) high mild
2 (2.00%) high severe
hex_binary/size=4096 time: [163.86 µs 164.18 µs 164.48 µs]
change: [−0.5476% −0.2517% +0.0122%] (p = 0.08 >
0.05)
No change in performance detected.
Found 3 outliers among 100 measurements (3.00%)
2 (2.00%) low mild
1 (1.00%) high mild
hex_binary/size=8192 time: [326.72 µs 327.45 µs 328.29 µs]
change: [−0.3658% −0.0302% +0.3973%] (p = 0.89 >
0.05)
No change in performance detected.
Found 5 outliers among 100 measurements (5.00%)
2 (2.00%) low mild
1 (1.00%) high mild
2 (2.00%) high severe
hex_int64_dict/size=1024
time: [6.6908 µs 6.8738 µs 7.0860 µs]
change: [−2.2874% −0.2044% +1.5752%] (p = 0.84 >
0.05)
No change in performance detected.
Found 8 outliers among 100 measurements (8.00%)
2 (2.00%) high mild
6 (6.00%) high severe
hex_int64_dict/size=4096
time: [27.934 µs 28.394 µs 28.926 µs]
change: [−1.8159% −0.5246% +0.7127%] (p = 0.43 >
0.05)
No change in performance detected.
Found 2 outliers among 100 measurements (2.00%)
2 (2.00%) high mild
hex_int64_dict/size=8192
time: [55.631 µs 56.239 µs 56.968 µs]
change: [−5.1281% −2.8229% −0.6794%] (p = 0.02 <
0.05)
Change within noise threshold.
Found 13 outliers among 100 measurements (13.00%)
5 (5.00%) high mild
8 (8.00%) high severe
```
</details>
I find the supposed int regressions surprising; the main hex loop was not
altered apart from simplifying how len is passed in. Maybe it was already so
fast that adding `make_scalar_function()`/altering the match path to get to the
hot loop is adding this extra overhead? 🤔
--
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]
