[connortsui20]

Summary

Progress towards #10020. Stacked on #10223.

A result can overflow the stored integer width while still fitting its logical type. Non-null bounds now select a sufficient width before checked arithmetic, so i8(127) + 1 can remain a narrow i64 result with i16 storage. If separate column extrema exceed the logical range, the existing checked kernel tests the actual row pairs at logical width.

Changes

Casts, fill, zip, between, concatenation, and interleave can use the stored children. Gather indices can also use their stored width directly. The rules retain logical dtypes, null behavior, and checked overflow. Buffers remain immutable, and the change does not add primitive integer types.

[codspeed]

Merging this PR will improve performance by 26.55%

⚠️ 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.

⚠️ 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

⚡ 26 improved benchmarks
✅ 2075 untouched benchmarks
🆕 3 new benchmarks
⏩ 518 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	fsst_decompress_string	526 µs	187.9 µs	×2.8
⚡	Simulation	chunked_canonicalize_into[(20, 100, 16, 4)]	739.9 µs	340.9 µs	×2.2
⚡	Simulation	chunked_dict_fsst_into_canonical[(250, 10, 2)]	681.9 µs	344.7 µs	+97.85%
⚡	Simulation	chunked_canonicalize_into[(20, 50, 8, 4)]	274.4 µs	213.4 µs	+28.58%
⚡	Simulation	chunked_canonicalize_into[(20, 50, 8, 16)]	279.5 µs	218.6 µs	+27.85%
⚡	Simulation	chunked_canonicalize_into[(20, 50, 8, 64)]	306.3 µs	243.2 µs	+25.98%
⚡	Simulation	chunked_into_canonical[(20, 50, 8, 4)]	314.3 µs	253 µs	+24.22%
⚡	Simulation	chunked_into_canonical[(20, 50, 8, 16)]	323.3 µs	262 µs	+23.4%
⚡	Simulation	chunked_into_canonical[(20, 50, 8, 64)]	348.2 µs	286.4 µs	+21.54%
⚡	Simulation	decompress_fsst[(200, 4, 4)]	60.8 µs	51 µs	+19.04%
⚡	Simulation	decompress_fsst[(200, 4, 8)]	61.6 µs	51.9 µs	+18.6%
⚡	Simulation	chunked_canonicalize_into[(10, 150, 4, 4)]	235.3 µs	199.5 µs	+17.94%
⚡	Simulation	decompress_fsst[(200, 16, 4)]	66.4 µs	56.7 µs	+17.13%
⚡	Simulation	decompress_fsst[(200, 16, 8)]	69.4 µs	59.8 µs	+15.98%
⚡	Simulation	chunked_canonicalize_into[(10, 150, 16, 4)]	266.6 µs	231 µs	+15.42%
⚡	Simulation	chunked_canonicalize_into[(20, 100, 16, 16)]	461.3 µs	400 µs	+15.33%
⚡	Simulation	chunked_into_canonical[(10, 150, 4, 4)]	264.9 µs	231 µs	+14.7%
⚡	Simulation	chunked_into_canonical[(20, 100, 16, 4)]	452.7 µs	398.1 µs	+13.72%
⚡	Simulation	chunked_into_canonical[(10, 150, 16, 4)]	298.7 µs	263.3 µs	+13.44%
⚡	Simulation	chunked_into_canonical[(20, 100, 16, 16)]	508.7 µs	448.6 µs	+13.41%
...	...	...	...	...	...

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

Tip

Curious why performance improved? Comment @codspeedbot explain why performance improved on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing ct/narrow-pushdown (557a613) with develop (3c87da0)²}

Footnotes

1. 518 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. ↩

2. No successful run was found on ct/narrow-encoding (44780e8) during the generation of this report, so develop (3c87da0) was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩