fastlanes: bit-packed compare-constant fast path + bitpack_constant kernel

[joseph-isaacs]

Summary

Stacked on #8012. Speeds up the bitpack_compare bench from the parent PR with two complementary optimizations driven by the same observation — a bit-packed lane holds values in [0, 2^bit_width - 1], so a constant outside that range can be answered analytically without touching the packed buffer.

Compare-constant fast path (compute/compare.rs)

Register a CompareKernel for BitPacked that short-circuits when the RHS constant c is outside [0, 2^bit_width - 1]. For each operator the answer is a constant boolean modulo patches and validity:

Operator	Outside range result
Eq / NotEq	false / true everywhere
Lt / Lte / Gt / Gte	constant once c is on either side of the range

Detecting the range is an O(1) i128 check via the new BitPackedData::value_fits_bit_width helper. With no patches and no nulls the kernel returns a ConstantArray<bool> (also O(1)); otherwise it allocates a BitBuffer, fills it with the constant result, and overlays the per-position outcome at each patch index. In-range constants fall through to the canonical decompress + Arrow compare path; tests exercise both fall-throughs.

bitpack_constant analytical encoder (array/bitpack_compress.rs)

Add a constant-only pack kernel that builds the FastLanes bit pattern for a [constant; len] input without calling BitPacking::pack. For constant input every lane produces the same bit_width output words; we compute those words analytically — each output word's j-th bit is bit (k * T_bits + j) mod bit_width of c — then memset each word LANES times into a stack chunk template and memcpy the template into every full chunk. The standard packer is only invoked for the partial tail (zero-padded past len). bitpack_encode_constant wraps the buffer up as a BitPackedArray. A bitwise-equivalence rstest covers byte-identity with BitPacking::pack across lengths, widths, and constants.

Benches

• bitpack_compare (added in bench: bit-packed compare-constant baseline #8012) on this branch now exercises the fast path; at bit_width ∈ {4, 16}, len ∈ {1024, 65536} it runs in ~1.4–1.5 µs vs 8–125 µs for the decompress + Arrow baseline.
• New bitpack_constant bench compares the analytical kernel against the full bitpack_encode pipeline on uniform-constant input; at 64 K u32 elements the analytical kernel is roughly 23–62× faster.

Plan doc (docs/inrange_compare_plan.md)

Documents the follow-up plan to accelerate in-range ordering comparisons: compare the packed array against the packed constant via SWAR less-than per supported bit width (Routes A/B/C, including Knuth broadword with rotation tables for widths that straddle word boundaries), derive the four ordering operators from one Lt primitive, and benchmark against the canonical SIMD baseline before landing.

Test plan

• cargo nextest run -p vortex-fastlanes --all-features → 265/265 pass locally
• cargo check -p vortex-fastlanes --benches --all-features
• cargo bench -p vortex-fastlanes --bench bitpack_compare shows the fast-path speedup vs the baseline from bench: bit-packed compare-constant baseline #8012
• cargo bench -p vortex-fastlanes --bench bitpack_constant shows the analytical encoder speedup
• ./scripts/public-api.sh agrees with the committed lock file
• cargo clippy --all-targets --all-features

Supersedes #8011 (split into bench + speedup).

🤖 Generated with Claude Code

[codspeed-hq]

Merging this PR will improve performance by 59.84%

⚠️ 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.

⚡ 9 improved benchmarks
❌ 5 regressed benchmarks
✅ 1261 untouched benchmarks
🆕 16 new benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	baseline_lt[16, 65536]	217.6 µs	303 µs	-28.19%
❌	Simulation	baseline_eq[4, 65536]	185 µs	243 µs	-23.86%
❌	Simulation	baseline_lt[4, 65536]	200.8 µs	258 µs	-22.17%
❌	Simulation	baseline_eq[16, 65536]	229.9 µs	287.8 µs	-20.13%
❌	Simulation	chunked_varbinview_canonical_into[(1000, 10)]	162.1 µs	198.2 µs	-18.2%
⚡	Simulation	fast_eq_out_of_range[16, 65536]	225.1 µs	36.2 µs	×6.2
⚡	Simulation	fast_lt_out_of_range[16, 65536]	207.8 µs	36.5 µs	×5.7
⚡	Simulation	fast_lt_out_of_range[4, 65536]	190.7 µs	35.8 µs	×5.3
⚡	Simulation	fast_eq_out_of_range[4, 65536]	174.6 µs	36.1 µs	×4.8
⚡	Simulation	fast_lt_out_of_range[4, 1024]	37.9 µs	26.5 µs	+42.93%
⚡	Simulation	baseline_lt[4, 1024]	78.9 µs	63.2 µs	+24.81%
⚡	Simulation	fast_lt_out_of_range[16, 1024]	31.8 µs	26.6 µs	+19.76%
⚡	Simulation	fast_eq_out_of_range[16, 1024]	32.1 µs	27.7 µs	+15.71%
⚡	Simulation	fast_eq_out_of_range[4, 1024]	31.4 µs	27.7 µs	+13.17%
🆕	Simulation	full_encode[16, 65536]	N/A	358 µs	N/A
🆕	Simulation	in_range_eq[4, 1024]	N/A	30.8 µs	N/A
🆕	Simulation	in_range_eq[16, 1024]	N/A	31 µs	N/A
🆕	Simulation	in_range_eq_baseline[4, 1024]	N/A	83.5 µs	N/A
🆕	Simulation	full_encode[4, 1024]	N/A	19.1 µs	N/A
🆕	Simulation	in_range_eq[16, 65536]	N/A	156.4 µs	N/A
...	...	...	...	...	...

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 claude/bitpack-compare-speedup-KGPS3 (940b84d) with develop (7de53da)¹}

Footnotes

1. No successful run was found on develop (41af74d) during the generation of this report, so 7de53da was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[github-actions]

This PR has been marked as stale because it has been open for 14 days with no activity. Please comment or remove the stale label if you wish to keep it active, otherwise it will be closed in 7 days

[github-actions]

This PR has been marked as stale because it has been open for 14 days with no activity. Please comment or remove the stale label if you wish to keep it active, otherwise it will be closed in 7 days

[github-actions]

This PR was closed because it has been inactive for 7 days since being marked as stale.