Potential regression induced by PR #56037

[rhshadrach]

PR #56037 may have induced a performance regression. If it was a necessary behavior change, this may have been expected and everything is okay.

Please check the links below. If any ASVs are parameterized, the combinations of parameters that a regression has been detected for appear as subbullets.

• inference.ToDatetimeFromIntsFloats.time_nanosec_float64
• inference.ToDatetimeFromIntsFloats.time_sec_float64

Subsequent benchmarks may have skipped some commits. The link below lists the commits that are between the two benchmark runs where the regression was identified.

05c32ba...d9f70b3

cc @jbrockmendel

[rtlee9]

My local asv benchmarking shows the regression was introduced in d9f70b3 rather than the previous commits

(venv) ➜  asv_bench git:(d9f70b397a) asv continuous -f 1.1 -E virtualenv HEAD~ HEAD -b inference.ToDatetimeFromIntsFloats.time_                                                                                                                                                                                                 Couldn't load asv.plugins._mamba_helpers because                                                                                                                                                                                                                                                                                No module named 'libmambapy'                                                                                                                                                                                                                                                                                                    · Creating environments                                                                                                                                                                                                                                                                                                         · Discovering benchmarks                                                                                                                                                                                                                                                                                                        ·· Uninstalling from virtualenv-py3.10-Cython3.0.5-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter.                                                                                                                                              ·· Installing d9f70b39 <v2.3.0.dev0~271> into virtualenv-py3.10-Cython3.0.5-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter.                                                                                                                     · Running 12 total benchmarks (2 commits * 1 environments * 6 benchmarks)                                                                                                                                                                                                                                                       [ 0.00%] · For pandas commit e37ff77b <v2.3.0.dev0~272> (round 1/2):                                                                                                                                                                                                                                                            [ 0.00%] ·· Building for virtualenv-py3.10-Cython3.0.5-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter..
[ 0.00%] ·· Benchmarking virtualenv-py3.10-Cython3.0.5-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter                                                                                                                                           [ 4.17%] ··· Running (inference.ToDatetimeFromIntsFloats.time_nanosec_float64--)......                                                                                                                                                                                                                                          [25.00%] · For pandas commit d9f70b39 <v2.3.0.dev0~271> (round 1/2):                                                                                                                                                                                                                                                            [25.00%] ·· Building for virtualenv-py3.10-Cython3.0.5-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter..
[25.00%] ·· Benchmarking virtualenv-py3.10-Cython3.0.5-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter
[29.17%] ··· Running (inference.ToDatetimeFromIntsFloats.time_nanosec_float64--)......                                                                                                                                                                                                                                          [50.00%] · For pandas commit d9f70b39 <v2.3.0.dev0~271> (round 2/2):                                                                                                                                                                                                                                                            [50.00%] ·· Benchmarking virtualenv-py3.10-Cython3.0.5-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter
[54.17%] ··· inference.ToDatetimeFromIntsFloats.time_nanosec_float64                                                                                 260±3ms
[58.33%] ··· inference.ToDatetimeFromIntsFloats.time_nanosec_int64                                                                               3.16±0.09ms
[62.50%] ··· inference.ToDatetimeFromIntsFloats.time_nanosec_uint64                                                                               3.03±0.2ms                                                                                                                                                                    [66.67%] ··· inference.ToDatetimeFromIntsFloats.time_sec_float64                                                                                     262±2ms                                                                                                                                                                    [70.83%] ··· inference.ToDatetimeFromIntsFloats.time_sec_int64                                                                                    31.1±0.3ms
[75.00%] ··· inference.ToDatetimeFromIntsFloats.time_sec_uint64                                                                                   30.9±0.2ms
[75.00%] · For pandas commit e37ff77b <v2.3.0.dev0~272> (round 2/2):
[75.00%] ·· Building for virtualenv-py3.10-Cython3.0.5-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter..
[75.00%] ·· Benchmarking virtualenv-py3.10-Cython3.0.5-jinja2-matplotlib-meson-meson-python-numba-numexpr-odfpy-openpyxl-pyarrow-python-build-scipy-sqlalchemy-tables-xlrd-xlsxwriter
[79.17%] ··· inference.ToDatetimeFromIntsFloats.time_nanosec_float64                                                                              6.17±0.3ms
[83.33%] ··· inference.ToDatetimeFromIntsFloats.time_nanosec_int64                                                                               2.93±0.06ms
[87.50%] ··· inference.ToDatetimeFromIntsFloats.time_nanosec_uint64                                                                              2.91±0.06ms
[91.67%] ··· inference.ToDatetimeFromIntsFloats.time_sec_float64                                                                                  5.51±0.3ms
[95.83%] ··· inference.ToDatetimeFromIntsFloats.time_sec_int64                                                                                    31.0±0.3ms
[100.00%] ··· inference.ToDatetimeFromIntsFloats.time_sec_uint64                                                                                  30.9±0.08ms
| Change   | Before [e37ff77b] <v2.3.0.dev0~272>   | After [d9f70b39] <v2.3.0.dev0~271>   |   Ratio | Benchmark (Parameter)                                   |
|----------|---------------------------------------|--------------------------------------|---------|---------------------------------------------------------|
| +        | 5.51±0.3ms                            | 262±2ms                              |   47.56 | inference.ToDatetimeFromIntsFloats.time_sec_float64     |
| +        | 6.17±0.3ms                            | 260±3ms                              |   42.2  | inference.ToDatetimeFromIntsFloats.time_nanosec_float64 |

SOME BENCHMARKS HAVE CHANGED SIGNIFICANTLY.
PERFORMANCE DECREASED.

[rhshadrach]

Thanks @rtlee9 - isn't that commit associated with the highlighted PR in the OP?

[rtlee9]

Yeah I was just confirming it was that commit in particular, since the asv benchmarks had skipped a few commits

Subsequent benchmarks may have skipped some commits. The link below lists the commits that are between the two benchmark runs where the regression was identified.

[rhshadrach]

Ah - thanks for confirming.