Use indexed sources for fallible constant rows

[connortsui20]

Summary

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

Extends the decoded input source from #9628 to fallible execute_owned calls.

Fallible RowFns already collect into spare Vec capacity while reducing compact failure evidence. Their constant fallback still selected between a row and a decoded constant through ElementTuple::get inside that loop, which hid the input layout from the optimizer and duplicated the traversal code used for ordinary row inputs.

Changes

Constructs one decoded source for every all-valid batch, then passes it to the existing map_checked_into collector. The source fixes each argument's row-or-constant choice before traversal, while the output allocation, initialization rules, and deferred failure handling remain unchanged.

This also removes the separate LaneZip execution branch. Optimized x86 IR keeps the same vector loop for varying Boolean inputs, unswitches source variants before the selected row loop, and removes the duplicate vectorized loop previously emitted for the two source representations. In the focused fixture, the deferred Boolean executor is 27% smaller and the deferred i64 executor is 19% smaller, with the same bounds-panic sites.

Tests cover a constant left-hand input. The focused benchmark covers varying inputs and both partial-constant orientations for deferred i64 output.

[codspeed-hq]

Merging this PR will degrade performance by 39.38%

⚠️ 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.

⚡ 7 improved benchmarks
❌ 23 regressed benchmarks
✅ 2100 untouched benchmarks
🆕 9 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	subtract_shapes_avx2[(16384, PerRowConstant)]	8.1 µs	20.1 µs	-59.62%
❌	WallTime	subtract_shapes_avx2[(16384, ConstantPerRow)]	8.2 µs	20.1 µs	-59.48%
❌	WallTime	sub_i64_constant_avx2	8.2 µs	20.1 µs	-59.39%
❌	WallTime	add_shapes_neon[(16384, PerRowConstant)]	9.8 µs	24.1 µs	-59.22%
❌	WallTime	add_i64_constant_neon	9.8 µs	24 µs	-59.08%
❌	WallTime	subtract_shapes_neon[(16384, PerRowConstant)]	10 µs	24.4 µs	-59.03%
❌	WallTime	sub_i64_constant_neon	10.1 µs	24.6 µs	-58.77%
❌	WallTime	add_shapes_neon[(16384, ConstantPerRow)]	9.9 µs	23.9 µs	-58.71%
❌	WallTime	subtract_shapes_neon[(16384, ConstantPerRow)]	10.1 µs	24.5 µs	-58.62%
❌	WallTime	mul_i32_constant_avx2	12 µs	28.7 µs	-58.32%
❌	WallTime	subtract_shapes_avx2[(16384, PerRowNullableConstant)]	8.7 µs	20.7 µs	-58.14%
❌	WallTime	add_shapes_neon[(16384, PerRowNullableConstant)]	10.3 µs	24.3 µs	-57.55%
❌	WallTime	subtract_shapes_neon[(16384, PerRowNullableConstant)]	10.7 µs	25 µs	-57.21%
❌	WallTime	mul_i32_constant_neon	11.5 µs	27 µs	-57.19%
❌	WallTime	add_i64_constant_avx2	8.1 µs	18.9 µs	-57.19%
❌	WallTime	add_shapes_avx2[(16384, PerRowConstant)]	8.1 µs	18.9 µs	-57.16%
❌	WallTime	add_shapes_avx2[(16384, ConstantPerRow)]	8.1 µs	18.9 µs	-57.1%
❌	WallTime	add_shapes_avx2[(16384, PerRowNullableConstant)]	8.6 µs	19.5 µs	-55.85%
❌	WallTime	scalar_subtract_neon	14.2 µs	21.1 µs	-32.59%
❌	WallTime	scalar_subtract_avx2	13.3 µs	18.9 µs	-29.9%
...	...	...	...	...	...

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-execute-owned-sources (95c4505) with ct/row-fn-constant-sources (d979a9b)}

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. ↩