Optimize list_contains with prepared constant sets

[robert3005]

Replace constant-set membership comparison/OR chains with prepared, single-pass probes

[codspeed]

Merging this PR will regress 10 benchmarks

⚠️ 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.

⚠️ 12 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

• pcodec_compress_f64
• case_when_fragmented[400]
• random_i128[0.01]
• density_sweep_dense_runs[0.001]
• multipolygon_mixed_validity
• filter_run_end[len=4096_run=16_density=0.1]
• ilike_contains
• decompress[u32, (1000, 256)]
• zstd_compress_string
• decompress[u16, (4000, 256)]
• …and 2 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

⚡ 5 improved benchmarks
❌ 10 regressed benchmarks
✅ 2168 untouched benchmarks
⏩ 359 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	optimize_lookup_predicate[ids=192, shape=in_list]	50.4 µs	75.5 µs	-33.2%
❌	Simulation	optimize_lookup_predicate[ids=128, shape=in_list]	51.1 µs	72.9 µs	-29.97%
❌	Simulation	optimize_lookup_predicate[ids=64, shape=in_list]	50.9 µs	70.3 µs	-27.6%
❌	WallTime	mul_u64_nonnull_neon	28.7 µs	39.2 µs	-26.64%
❌	Simulation	optimize_lookup_predicate[ids=16, shape=in_list]	51.7 µs	69.1 µs	-25.1%
❌	Simulation	optimize_lookup_predicate[ids=256, shape=in_list]	79.4 µs	104.7 µs	-24.2%
❌	Simulation	optimize_lookup_predicate[ids=1, shape=in_list]	74.8 µs	89.6 µs	-16.46%
❌	WallTime	multiply_shapes_neon[(32768, PerRowPerRow)]	32.7 µs	38.1 µs	-14.15%
❌	WallTime	mul_i64_nonnull_neon	32.7 µs	38.1 µs	-14.08%
❌	Simulation	or_true_constant	18.5 µs	20.6 µs	-10.2%
⚡	Simulation	i64_random_chunked[256]	15,817.7 µs	171.1 µs	×92
⚡	Simulation	utf8_random[256]	13,524.8 µs	355.5 µs	×38
⚡	Simulation	i64_random[256]	4,360.6 µs	131.1 µs	×33
⚡	Simulation	nested_list_random[32]	6.7 ms	1 ms	×6.7
⚡	WallTime	bitpack_blocked_compress_avx512	5.6 µs	4.7 µs	+18.98%

Tip

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

---

_{Comparing claude/friendly-knuth-dj6ed1-list-contains (79947bc) with develop (731a231)}

Footnotes

1. 359 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[joseph-isaacs]

How do i extend this to work for new custom kernels?

I really think we want to make probe an array to allow reduce/execute rules?

[robert3005]

this is a hack, we essentially have common setup we would like to share across chunks, if we just dispatch over each chunk there's no place to store the shared state