Micha Reiser
|
40f54375cb
|
Pull in RustPython parser (#6099)
|
2023-07-27 09:29:11 +00:00 |
|
Micha Reiser
|
2cf00fee96
|
Remove parser dependency from ruff-python-ast (#6096)
|
2023-07-26 17:47:22 +02:00 |
|
Charlie Marsh
|
ed72c027a3
|
Replace NoHashHasher usages with FxHashMap (#6049)
## Summary
I had always assumed that `NoHashHasher` would be faster when using
integer keys, but benchmarking shows otherwise:
```
linter/default-rules/numpy/globals.py
time: [66.544 µs 66.606 µs 66.678 µs]
thrpt: [44.253 MiB/s 44.300 MiB/s 44.342 MiB/s]
change:
time: [-0.1843% +0.1087% +0.3718%] (p = 0.46 > 0.05)
thrpt: [-0.3704% -0.1086% +0.1847%]
No change in performance detected.
Found 1 outliers among 100 measurements (1.00%)
1 (1.00%) high mild
linter/default-rules/pydantic/types.py
time: [1.3787 ms 1.3811 ms 1.3837 ms]
thrpt: [18.431 MiB/s 18.466 MiB/s 18.498 MiB/s]
change:
time: [-0.4827% -0.1074% +0.1927%] (p = 0.56 > 0.05)
thrpt: [-0.1924% +0.1075% +0.4850%]
No change in performance detected.
linter/default-rules/numpy/ctypeslib.py
time: [624.82 µs 625.96 µs 627.17 µs]
thrpt: [26.550 MiB/s 26.601 MiB/s 26.650 MiB/s]
change:
time: [-0.7071% -0.4908% -0.2736%] (p = 0.00 < 0.05)
thrpt: [+0.2744% +0.4932% +0.7122%]
Change within noise threshold.
linter/default-rules/large/dataset.py
time: [3.1585 ms 3.1634 ms 3.1685 ms]
thrpt: [12.840 MiB/s 12.861 MiB/s 12.880 MiB/s]
change:
time: [-1.5338% -1.3463% -1.1476%] (p = 0.00 < 0.05)
thrpt: [+1.1610% +1.3647% +1.5577%]
Performance has improved.
linter/all-rules/numpy/globals.py
time: [140.17 µs 140.37 µs 140.58 µs]
thrpt: [20.989 MiB/s 21.020 MiB/s 21.051 MiB/s]
change:
time: [-0.1066% +0.3140% +0.7479%] (p = 0.14 > 0.05)
thrpt: [-0.7423% -0.3130% +0.1067%]
No change in performance detected.
Found 3 outliers among 100 measurements (3.00%)
2 (2.00%) high mild
1 (1.00%) high severe
linter/all-rules/pydantic/types.py
time: [2.7030 ms 2.7069 ms 2.7112 ms]
thrpt: [9.4064 MiB/s 9.4216 MiB/s 9.4351 MiB/s]
change:
time: [-0.6721% -0.4874% -0.2974%] (p = 0.00 < 0.05)
thrpt: [+0.2982% +0.4898% +0.6766%]
Change within noise threshold.
Found 14 outliers among 100 measurements (14.00%)
12 (12.00%) high mild
2 (2.00%) high severe
linter/all-rules/numpy/ctypeslib.py
time: [1.4709 ms 1.4727 ms 1.4749 ms]
thrpt: [11.290 MiB/s 11.306 MiB/s 11.320 MiB/s]
change:
time: [-1.1617% -0.9766% -0.8094%] (p = 0.00 < 0.05)
thrpt: [+0.8160% +0.9862% +1.1754%]
Change within noise threshold.
Found 12 outliers among 100 measurements (12.00%)
9 (9.00%) high mild
3 (3.00%) high severe
linter/all-rules/large/dataset.py
time: [5.8086 ms 5.8163 ms 5.8240 ms]
thrpt: [6.9854 MiB/s 6.9946 MiB/s 7.0038 MiB/s]
change:
time: [-1.5651% -1.3536% -1.1584%] (p = 0.00 < 0.05)
thrpt: [+1.1720% +1.3721% +1.5900%]
Performance has improved.
```
My guess is that `NoHashHasher` underperforms because the keys are not
randomly distributed...
Anyway, it's a ~1% (significant) performance gain on some of the above,
plus we get to remove a dependency.
|
2023-07-24 23:41:57 +00:00 |
|
Charlie Marsh
|
68b6d30c46
|
Use consistent Cargo.toml metadata in all crates (#5015)
|
2023-06-12 00:02:40 +00:00 |
|
Charlie Marsh
|
ea31229be0
|
Track TYPE_CHECKING blocks in Importer (#4593)
|
2023-05-30 16:18:10 +00:00 |
|
Micha Reiser
|
652c644c2a
|
Introduce ruff_index crate (#4597)
|
2023-05-23 17:40:35 +02:00 |
|
Micha Reiser
|
cab65b25da
|
Replace row/column based Location with byte-offsets. (#3931)
|
2023-04-26 18:11:02 +00:00 |
|
Charlie Marsh
|
d919adc13c
|
Introduce a ruff_python_semantic crate (#3865)
|
2023-04-04 16:50:47 +00:00 |
|