Pack deferred Boolean RowFn output directly

[connortsui20]

Summary

• Tracking Issue: Epic: Row-oriented scalar functions #9128

Extends direct output collection to the all-valid fallible execute_owned path. Depends on #9629.

The source work in the preceding PRs makes each input layout visible to the collector, but fallible Boolean RowFns still write one byte per result into Vec<bool> before constructing a packed array. The output type needs to own collection while the executor also reduces failure evidence.

Changes

Adds OutputElement::build_from_deferred, whose default keeps the existing Vec and map_checked_into collection. The Boolean override packs output with BitBuffer::collect_bool while OR-reducing the matching failure evidence. execute_owned passes its decoded source through that hook.

Tests cover varying inputs, both partial-constant orientations, and an observable valid-row failure. The focused benchmark covers deferred Boolean output for i32 and i64, with deferred i64 as the non-Boolean control. Nullable retry and valid-only execution are unchanged.

Optimized x86 codegen emits vector comparisons, vpmovmskb, and direct packed-word stores for varying inputs and both constant orientations. Deferred i64 remains scalar, so this stays a draft experiment rather than evidence that the generic fallible path improved.

The hook constructs the ArrayRef before finish_failure validates the reduced evidence. A rejected batch therefore pays output-construction cost and discards the column, and a custom OutputElement observes output-construction panics before the deferred error is built.

API Changes

Adds OutputElement::build_from_deferred. Implementations must preserve the row order and output contract of build_from, return the bitwise OR of exactly the callback failure values, and cannot add another failure path. Failure evidence must be no wider than the output element.

[codspeed-hq]

Merging this PR will degrade performance by 46.65%

⚠️ 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

⚡ 6 improved benchmarks
❌ 31 regressed benchmarks
✅ 2075 untouched benchmarks
🆕 45 new benchmarks
⏩ 106 skipped benchmarks¹
🗄️ 4 archived benchmarks run²

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	WallTime	mul_i32_constant_avx512	7.9 µs	28.8 µs	-72.54%
❌	WallTime	subtract_shapes_avx512[(16384, ConstantPerRow)]	7 µs	20.2 µs	-65.21%
❌	WallTime	sub_i64_constant_avx512	7.1 µs	20.1 µs	-64.71%
❌	WallTime	subtract_shapes_avx512[(16384, PerRowConstant)]	7.1 µs	20.1 µs	-64.66%
❌	WallTime	add_shapes_avx512[(16384, PerRowConstant)]	7 µs	19 µs	-63.15%
❌	WallTime	add_i64_constant_avx512	7 µs	19 µs	-63.07%
❌	WallTime	subtract_shapes_avx512[(16384, PerRowNullableConstant)]	7.6 µs	20.7 µs	-63.06%
❌	WallTime	add_shapes_avx512[(16384, ConstantPerRow)]	7.1 µs	19 µs	-62.86%
❌	WallTime	add_shapes_avx512[(16384, PerRowNullableConstant)]	7.6 µs	19.6 µs	-61.19%
❌	WallTime	subtract_shapes_avx2[(16384, PerRowConstant)]	8.1 µs	20.1 µs	-59.8%
❌	WallTime	sub_i64_constant_avx2	8.1 µs	20.2 µs	-59.78%
❌	WallTime	subtract_shapes_avx2[(16384, ConstantPerRow)]	8.1 µs	20.2 µs	-59.71%
❌	WallTime	add_shapes_neon[(16384, ConstantPerRow)]	9.7 µs	23.9 µs	-59.3%
❌	WallTime	add_shapes_neon[(16384, PerRowConstant)]	9.8 µs	23.9 µs	-59.21%
❌	WallTime	add_i64_constant_neon	9.8 µs	24 µs	-58.95%
❌	WallTime	subtract_shapes_neon[(16384, PerRowConstant)]	10 µs	24.3 µs	-58.74%
❌	WallTime	subtract_shapes_avx2[(16384, PerRowNullableConstant)]	8.6 µs	20.7 µs	-58.37%
❌	WallTime	mul_i32_constant_avx2	12 µs	28.8 µs	-58.36%
❌	WallTime	add_shapes_neon[(16384, PerRowNullableConstant)]	10.5 µs	24.8 µs	-57.86%
❌	WallTime	sub_i64_constant_neon	10.4 µs	24.6 µs	-57.73%
...	...	...	...	...	...

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 ct/row-fn-bool-output-deferred (4d96e40) with develop (ee1ac25)³}

Footnotes

1. 106 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. 4 benchmarks were run, but are now archived. If they were deleted in another branch, consider rebasing to remove them from the report. Instead if they were added back, click here to restore them. ↩

3. No successful run was found on ct/row-fn-execute-owned-sources (95c4505) during the generation of this report, so develop (ee1ac25) was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩