perf(python): release the GIL once per scan export

[robert3005]

Scans that export to Arrow from Python release the GIL fewer times. When other threads hold the GIL, each release costs a wait to take it back.

[codspeed]

Merging this PR will degrade performance by 38.33%

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

⚠️ 1 benchmark measured no execution time

Nothing ran under measurement, usually because the compiler removed the code under test. This result is not comparable, so it counts as unchanged.

Preventing compiler optimizations

• bench_compare_primitive[(10000, 2)]

⚠️ 17 benchmarks spent significant time in system calls

System calls cannot be consistently instrumented, so they are not included in the measure, which understates the real cost. Please switch to the Walltime instrument to accurately measure system calls.

Measurement and system calls

• density_sweep_dense_runs[0.001]
• random_i128[0.01]
• take_varbin
• test_scan_filter[int_str_float]
• decompress[u32, (1000, 256)]
• from_vec_drop_vortex[65536]
• test_scan[int_str]
• test_scan[int_str_float]
• test_scan_filter[int_float]
• decompress[u16, (4000, 256)]
• …and 7 more

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 15 improved benchmarks
❌ 6 regressed benchmarks
✅ 2222 untouched benchmarks
🆕 16 new benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	i64_random_chunked[256]	226.6 µs	13,908.3 µs	-98.37%
❌	Simulation	utf8_random[256]	416.6 µs	12,969.1 µs	-96.79%
❌	Simulation	i64_random[256]	170.2 µs	3,941.4 µs	-95.68%
❌	Simulation	nested_list_random[32]	1 ms	6.7 ms	-84.48%
❌ ⚠️	Simulation	density_sweep_dense_runs[0.001]	28.7 µs	47.6 µs	-39.8%
❌	WallTime	threshold_avx2[8]	1.9 µs	2.2 µs	-14.16%
⚡	Simulation	optimize_lookup_predicate[ids=192, shape=in_list]	76.4 µs	51 µs	+49.81%
⚡	Simulation	optimize_lookup_predicate[ids=128, shape=in_list]	73.8 µs	51.7 µs	+42.76%
⚡	Simulation	optimize_lookup_predicate[ids=64, shape=in_list]	71.2 µs	51.5 µs	+38.15%
⚡	Simulation	optimize_lookup_predicate[ids=16, shape=in_list]	70 µs	52.4 µs	+33.66%
⚡	Simulation	optimize_lookup_predicate[ids=256, shape=in_list]	105.9 µs	79.6 µs	+32.93%
⚡	WallTime	threshold_neon[8]	3.6 µs	2.9 µs	+23.65%
⚡	Simulation	optimize_lookup_predicate[ids=1, shape=in_list]	90.5 µs	75.3 µs	+20.26%
⚡	Simulation	sum_v2_i64	224.8 µs	194.5 µs	+15.6%
⚡	Simulation	sum_i64	224.6 µs	194.4 µs	+15.55%
⚡	Simulation	random_i8[0.01]	124.1 µs	108.6 µs	+14.29%
⚡	WallTime	starts_with_avx512[8, VarBin]	46.4 µs	40.7 µs	+14.19%
⚡	Simulation	test_scan_filter[int]	3.5 ms	3.1 ms	+13.44%
⚡	WallTime	filter_neon[80]	3.5 µs	3.1 µs	+13.02%
⚡ ⚠️	Simulation	random_i128[0.01]	8.1 µs	7.2 µs	+11.68%
...	...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing rk/python-gil (40cfa27) with rk/python-codspeed (029e60c)¹}

Footnotes

1. No successful run was found on rk/python-codspeed (a135e85) during the generation of this report, so 8bdcedc was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[joseph-isaacs]

Shall we add py benchmarks?