perf: specialize grouped sums for constant arrays

[connortsui20]

Summary

Reuses one sum partial for fixed-size lists of primitive constant elements. Variable-size groups reuse the existing whole-array accumulator without slicing the elements.

Changes

Uses the existing Sum and SumV2 accumulation paths for nulls, NaNs, overflow, and signed zero, with focused tests for grouped behavior. Whole-array arithmetic and decimal handling remain unchanged.

Validation

• cargo nextest run -p vortex-array --lib: 3,462 passed, 1 skipped.
• cargo test --doc -p vortex-array: passed.
• cargo clippy --all-targets --all-features -- -D warnings: passed.
• Affected-crate Rust formatting passed. Full workspace formatting still reports four existing macro-format differences in vortex-duckdb and vortex-ffi, outside this PR.

Benchmark results

Local aarch64-apple-darwin medians with Rust 1.98.0 and 100 samples, using cargo bench -p vortex-array --bench aggregate_grouped -- sum_v2_constant --sample-count 100. Both modes use f185f493f4be4568e82d75fb8869f460e25e1c0f, with the baseline disabling the constant grouped kernel. Timings include finalization to a primitive array.

Fixed-size inputs contain 16,384 i32 values. The list-view input has 128 groups with sizes from 1 to 15.

Workload	Specialized	Kernel disabled
Fixed-size, 2-element groups	4.541 µs	5.646 ms
Fixed-size, 128-element groups	3.999 µs	87.79 µs
Variable-size list groups	70.97 µs	87.02 µs

[codspeed-hq]

Merging this PR will regress 3 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.

⚠️ 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
❌ 3 regressed benchmarks
✅ 2184 untouched benchmarks
🆕 6 new benchmarks
⏩ 218 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	WallTime	arrow_checked_add_u32_neon[16384]	13.4 µs	20.3 µs	-34.07%
❌	Simulation	decompress[u64, (4000, 1024)]	70.8 µs	86 µs	-17.7%
❌	WallTime	mul_u32_nonnull_avx512	5.5 µs	6.2 µs	-10.08%
⚡	WallTime	mul_u64_nonnull_neon	20.6 µs	15.1 µs	+36.24%
⚡	WallTime	multiply_shapes_neon[(16384, PerRowPerRow)]	20.3 µs	17.2 µs	+17.98%
⚡	WallTime	mul_i64_nonnull_neon	20.2 µs	17.2 µs	+17.71%
⚡	WallTime	filtered_owned_i64_avx512[OneNullInEight]	26.3 µs	22.4 µs	+17.12%
⚡	Simulation	allocate_drop_arrow[0]	456.9 ns	402.7 ns	+13.45%
⚡	WallTime	mul_i32_nonnull_avx512	7.9 µs	7.1 µs	+10.34%
🆕	Simulation	sum_v2_constant_fixed[false, 128]	N/A	1.4 ms	N/A
🆕	Simulation	sum_v2_constant_fixed[false, 2]	N/A	82.6 ms	N/A
🆕	Simulation	sum_v2_constant_fixed[true, 128]	N/A	168 µs	N/A
🆕	Simulation	sum_v2_constant_fixed[true, 2]	N/A	229 µs	N/A
🆕	Simulation	sum_v2_constant_list[false]	N/A	1.4 ms	N/A
🆕	Simulation	sum_v2_constant_list[true]	N/A	1.2 ms	N/A

Tip

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

---

_{Comparing ct/sum-constant (f185f49) with develop (e3b8eb2)}

Footnotes

1. 218 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. ↩