Remove unsafe code from compute_vec_l2sq

[arrayka]

Summary

Removes the unsafe { std::slice::from_raw_parts(...) } call from compute_vec_l2sq by replacing manual index arithmetic with chunks_exact / par_chunks_exact iterators zipped with the output slice.

Changes

• compute_vec_l2sq: Now takes &[f32] directly instead of (data, index, dim). No more unsafe pointer arithmetic.
• compute_vecs_l2sq: Uses zip(data.chunks_exact(dim)) (sequential path) and zip(data.par_chunks_exact(dim)) (parallel path) to feed slices into compute_vec_l2sq. Eliminates enumerate() and manual offset calculation.

Benchmark Results

Benchmarks are flat or within noise threshold.

iai-callgrind (snrm2_benchmark_rust_iai — directly benchmarks compute_vecs_l2sq) shows small perf improvement (# of CPU instructions reduced by 2%):

cargo bench -p diskann-disk --bench bench_main_iai
    Finished `bench` profile [optimized + debuginfo] target(s) in 24.13s
     Running benches/bench_main_iai.rs (target/release/deps/bench_main_iai-f51a979b5ebb65bd)
bench_main_iai::kmeans_bench_iai::snrm2_benchmark_rust_iai
  Instructions:                     2371584|2421508              (-2.06169%) [-1.02105x]
  L1 Hits:                          2925062|2975035              (-1.67974%) [-1.01708x]
  L2 Hits:                            51229|51167                (+0.12117%) [+1.00121x]
  RAM Hits:                           32950|32947                (+0.00911%) [+1.00009x]
  Total read+write:                 3009241|3059149              (-1.63143%) [-1.01658x]
  Estimated Cycles:                 4334457|4384015              (-1.13042%) [-1.01143x]

Iai-Callgrind result: Ok. 3 without regressions; 0 regressed; 3 benchmarks finished in 94.0474s

Criterion (Snrm2 Rust Run): No statistically significant change (p > 0.05 in both before/after runs):

.\bench_main-8c2599eee32d59ce_after.exe --bench Snrm2 --color=never
Gnuplot not found, using plotters backend
Benchmarking kmeans-computation/Snrm2 Rust Run
Benchmarking kmeans-computation/Snrm2 Rust Run: Warming up for 3.0000 s
Benchmarking kmeans-computation/Snrm2 Rust Run: Collecting 50 samples in estimated 5.9923 s (7650 iterations)
Benchmarking kmeans-computation/Snrm2 Rust Run: Analyzing
kmeans-computation/Snrm2 Rust Run
                        time:   [777.50 µs 787.89 µs 799.10 µs]
                        change: [-3.9665% -2.4543% -0.7324%] (p = 0.00 < 0.05)
                        Change within noise threshold.
Found 2 outliers among 50 measurements (4.00%)
  2 (4.00%) high mild

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.48%. Comparing base (895a2c0) to head (5f09e10).
⚠️ Report is 7 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1094      +/-   ##
==========================================
+ Coverage   89.45%   89.48%   +0.03%     
==========================================
  Files         458      474      +16     
  Lines       85398    89753    +4355     
==========================================
+ Hits        76392    80317    +3925     
- Misses       9006     9436     +430     

Flag	Coverage Δ
miri	89.48% <100.00%> (+0.03%)	⬆️
unittests	89.13% <100.00%> (+0.05%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
diskann-disk/src/utils/math_util.rs	98.83% <100.00%> (+0.02%)	⬆️

... and 46 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

This PR refactors compute_vec_l2sq/compute_vecs_l2sq in diskann-disk to remove unsafe slice construction by switching from manual pointer/index arithmetic to safe slice chunk iterators (sequential and Rayon-parallel), while preserving performance characteristics.

Changes:

• Refactors compute_vec_l2sq to accept a &[f32] vector slice directly (removing unsafe { from_raw_parts(...) }).
• Updates compute_vecs_l2sq to drive computation via chunks_exact(dim) and par_chunks_exact(dim) zipped with the output buffer.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.