PiPNN 2/6: add numerical kernels

[weiyaoluo (SeliMeli)]

Summary

Add GEMM-backed distance computation and SIMD top-k selection for PiPNN behind the opt-in diskann/pipnn feature.

The numerical layer supports two operations:

• Partition assignment: rank sampled leaders for each point and return their local IDs.
• Leaf neighbor selection: compute each unordered point pair once and update both endpoints’ neighbor lists, returning local IDs and distances.

This is part 2 of the PiPNN stack. Recursive graph construction integrates these kernels in #1290.

Design

Separate metric computation from neighbor selection. Metric implementations handle L2, cosine, normalized cosine, and inner product. Partition leaders retain metric-specific norms for reuse across point stripes. Kernels choose which distance slices
to process; the shared TopK implementation handles SIMD loading, threshold filtering, scalar tails, and sorted insertion.

Reuse each leaf distance for both endpoints. New diskann-linalg helpers compute or accumulate the lower triangle of a Gram matrix using faer. Leaf selection scans the strict lower triangle, excludes self-neighbors, and shares each loaded distance
block between both endpoint updates.

Specialize common top-k capacities. Capacities 1–10 use compile-time specialization to enable insertion-loop unrolling; larger capacities use the runtime path. Typical partition fanouts are 10 and 3, and typical leaf-neighbor counts are 2 or 3.
Architecture-specific selection runs through diskann-wide at the TopK entry points.

Keep storage reusable. TopK borrows caller-owned candidate buffers. Kernels manage distance and ranking scratch without exposing SIMD block types to their callers.

The linear-algebra helpers validate matrix shapes and size-product overflow.

[Copilot]

Pull request overview

This PR adds the first set of PiPNN “kernel” building blocks to the DiskANN Rust workspace: SIMD-accelerated top‑k selection for partition assignment and leaf neighbor selection, along with supporting SIMD division and a new lower-triangular A·Aᵀ helper in diskann-linalg.

Changes:

• Add a new diskann-pipnn crate with partition_kernel and leaf_kernel implementations plus extensive correctness tests and Criterion benchmarks.
• Extend diskann-wide to support Div on relevant f32 SIMD types (native, doubled, and scalar/emulated) and add a corresponding division test macro.
• Add diskann_linalg::sgemm_aat_lower (lower-triangle-only AAT) and wire new crate/tests/CI/mutants exclusions into the workspace.

Reviewed changes

Copilot reviewed 26 out of 27 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
diskann-wide/src/test_utils/ops.rs	Adds test_div! macro to validate lane-wise SIMD division correctness.
diskann-wide/src/emulated.rs	Adds Div for scalar/emulated Emulated<f32, N, A> to support division in scalar dispatch.
diskann-wide/src/doubled.rs	Adds Div for Doubled<T> to support composite SIMD widths.
diskann-wide/src/arch/x86_64/v4/f32x8_.rs	Adds AVX Div op mapping + division tests.
diskann-wide/src/arch/x86_64/v4/f32x4_.rs	Adds SSE Div op mapping + division tests.
diskann-wide/src/arch/x86_64/v4/f32x16_.rs	Adds AVX-512 Div op mapping + division tests.
diskann-wide/src/arch/x86_64/v3/f32x8_.rs	Adds AVX Div op mapping + division tests for V3.
diskann-wide/src/arch/x86_64/v3/f32x4_.rs	Adds SSE Div op mapping + division tests for V3.
diskann-wide/src/arch/x86_64/v3/f32x16_.rs	Adds division tests for the f32x16 V3 path (likely via doubled composition).
diskann-wide/src/arch/aarch64/f32x4_.rs	Adds Neon Div op mapping + division tests.
diskann-wide/src/arch/aarch64/f32x2_.rs	Adds Neon Div op mapping + division tests.
diskann-pipnn/tests/partition_kernel.rs	New integration tests for partition top‑k dispatch correctness and edge cases.
diskann-pipnn/tests/leaf_kernel.rs	New integration tests for leaf neighbor top‑k dispatch correctness and edge cases.
diskann-pipnn/src/partition_kernel/tests.rs	New unit tests comparing scalar reference vs runtime dispatch and metric contracts.
diskann-pipnn/src/partition_kernel.rs	New partition-assignment distance + top‑k kernel with validation and SIMD dispatch.
diskann-pipnn/src/lib.rs	New crate root exporting PiPNN kernel modules.
diskann-pipnn/src/leaf_kernel/tests.rs	New unit tests for scalar reference parity and workspace behavior.
diskann-pipnn/src/leaf_kernel.rs	New fused lower-triangle leaf neighbor kernel with SIMD dispatch and workspace support.
diskann-pipnn/Cargo.toml	Defines new diskann-pipnn crate, dev-deps, and benches.
diskann-pipnn/benches/kernels.rs	Adds benchmarks for partition top‑k, lower AAT, leaf top‑k, and full leaf workflow.
diskann-linalg/tests/sgemm_aat_lower.rs	New tests for lower-triangle AAT behavior and validation errors.
diskann-linalg/src/lib.rs	Adds public sgemm_aat_lower API with dimension checks.
diskann-linalg/src/faer.rs	Implements sgemm_aat_lower_impl using Faer triangular matmul.
Cargo.toml	Adds diskann-pipnn to workspace members and workspace dependencies.
Cargo.lock	Records the new diskann-pipnn package entry.
.github/workflows/ci.yml	Adds diskann-pipnn to CI test package lists.
.cargo/mutants.toml	Adds mutation-test exclusions for kernel code paths and equivalent transformations.

Comments suppressed due to low confidence (2)

diskann-pipnn/src/leaf_kernel.rs:651

• Same issue as the L2 arm: using max_simd for lower clamping can erase NaNs on the Scalar/Emulated backend, making NaN distances rankable. Clamp with lt_simd + select to preserve NaNs consistently.

        Metric::CosineNormalized => {
            let distance = F::splat(arch, 1.0) - dot;
            zero.max_simd(distance)
        }

diskann-pipnn/src/leaf_kernel.rs:664

• The cosine path also uses zero.max_simd(distance) for clamping, which can collapse NaNs to zero on the Scalar/Emulated backend (via f32::max). That contradicts the comment about preserving non-rankable NaNs and can change output ordering. Prefer an lt_simd + select clamp here as well.

            let distance = one - cosine;
            // Comparisons with NaN are false, so this explicit lower clamp
            // preserves non-rankable NaNs while matching the existing PiPNN
            // distance formulas for finite values.
            zero.max_simd(distance)

---

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

[Copilot]

Pull request overview

Copilot reviewed 26 out of 27 changed files in this pull request and generated no new comments.

[Codecov Comments Bot (codecov-commenter)]

Codecov Report

❌ Patch coverage is 98.11747% with 25 lines in your changes missing coverage. Please review.
✅ Project coverage is 91.61%. Comparing base (600c2b9) to head (6779f44).
⚠️ Report is 33 commits behind head on main.

Files with missing lines	Patch %	Lines
diskann/src/graph/pipnn/topk.rs	98.65%	6 Missing ⚠️
diskann/src/graph/pipnn/leaf_metric.rs	95.34%	4 Missing ⚠️
diskann/src/graph/pipnn/partition_kernel.rs	97.29%	4 Missing ⚠️
diskann-linalg/src/lib.rs	98.63%	3 Missing ⚠️
diskann/src/graph/pipnn/leaf_kernel.rs	97.79%	3 Missing ⚠️
diskann/src/graph/pipnn/partition_metric.rs	97.97%	3 Missing ⚠️
diskann/src/graph/pipnn/mod.rs	98.21%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1287      +/-   ##
==========================================
+ Coverage   91.55%   91.61%   +0.06%     
==========================================
  Files         521      567      +46     
  Lines      100302   112317   +12015     
==========================================
+ Hits        91828   102904   +11076     
- Misses       8474     9413     +939     

Flag	Coverage Δ
miri	91.61% <98.11%> (+0.06%)	⬆️
unittests	91.56% <98.11%> (+0.33%)	⬆️

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

Files with missing lines	Coverage Δ
diskann-linalg/src/faer.rs	100.00% <100.00%> (ø)
diskann/src/graph/pipnn/simd.rs	100.00% <100.00%> (ø)
diskann/src/graph/pipnn/mod.rs	98.21% <98.21%> (ø)
diskann-linalg/src/lib.rs	99.20% <98.63%> (-0.49%)	⬇️
diskann/src/graph/pipnn/leaf_kernel.rs	97.79% <97.79%> (ø)
diskann/src/graph/pipnn/partition_metric.rs	97.97% <97.97%> (ø)
diskann/src/graph/pipnn/leaf_metric.rs	95.34% <95.34%> (ø)
diskann/src/graph/pipnn/partition_kernel.rs	97.29% <97.29%> (ø)
diskann/src/graph/pipnn/topk.rs	98.65% <98.65%> (ø)

... and 253 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

Copilot reviewed 26 out of 27 changed files in this pull request and generated no new comments.

Comments suppressed due to low confidence (2)

diskann-pipnn/src/partition_kernel/tests.rs:20

• The PartitionTopK contract for Metric::L2 expects leader_scales to contain squared leader norms (see docs and distance(Metric::L2, ..) test). This helper currently populates unsquared norms, which makes the test data inconsistent with the public API contract and could hide contract-related bugs.

    let leader_scales = match metric {
        Metric::L2 => (0..leaders).map(|leader| (leader + 1) as f32).collect(),
        Metric::Cosine => (0..leaders)
            .map(|leader| {

diskann-pipnn/src/partition_kernel.rs:61

• InvalidFanout’s error message says the maximum is {maximum}, but validation also rejects fanout > leaders. When leaders < maximum this message is misleading (it implies the only limit is {maximum}). Consider spelling out both constraints in the message so callers immediately see why it failed.

    #[error("invalid fanout {fanout} for {leaders} leaders; maximum is {maximum}")]

[Copilot]

Pull request overview

Copilot reviewed 25 out of 26 changed files in this pull request and generated no new comments.

Suppressed comments (1)

diskann-pipnn/src/partition_kernel.rs:294

• For Metric::Cosine, NaN norms currently produce a finite distance (1.0) because denominator.gt_simd(0) is false for NaN, so the lane falls back to cosine = 0. That makes NaN-derived pairs/leaders “rankable”, which contradicts the module’s stated NaN-rejection behavior and differs from diskann-vector cosine semantics (NaN norms propagate to a NaN similarity/distance). Consider explicitly preserving NaN denominators so the resulting distance stays NaN and is ignored by insert_topk.

        let denominator = row_norm * leader_norm;
        let valid = denominator.gt_simd(zero);
        let safe_denominator = valid.select(denominator, one);
        let cosine = valid.select(dot / safe_denominator, zero);
        one - cosine

[juchen-ms (partychen)]

Nice work overall. I found one correctness issue in the cosine handling that should be resolved before merge. The remaining comments are mostly about reducing duplicated or unsafe code and tightening the API contracts.

[Aditya Krishnan (arkrishn94)]

Thanks Weiyao, this is progress from the previous mega-PR. I still have some big-picture comments (we covered most of these offline) -

• Documentation: As I mentioned, we need thorough documentation in the diskann-pipnn crate. The main modules, partition_kernel and leaf_kernel need documentation up top, highlighting the main structures and how they are used - e.g. process_rows_binary/unary and nearest_leaders. Similarly with process_pairs_simd_* and nearest_leaf_neighbors
• Testing: I am concerned about the lack of testing for partition_kernel.rs and leaf_kernel.rs.
  • I notice some e2e integration tests but these kernels should be thoroughly tested, sweeping different input parameters, architectures and edge cases. This is especially needed given the amount of unsafe code.
  • That brings me to miri - there should be miri tests too.
  • I'm curious why are the tests in a separate submodule to the main files (for partition_kernel.rs and leaf_kernel.rs)? Let's try to keep tests along with the code being tested.
• Criterion: Since criterion is not a standard part of our library for benchmarking, let us not introduce it for this crate.
• Kernel dispatch: I left comments about you're disptaching the kernels, please take a look.

[Copilot]

Pull request overview

Copilot reviewed 26 out of 27 changed files in this pull request and generated no new comments.

Suppressed comments (2)

diskann-pipnn/src/partition_kernel/tests.rs:19

• PartitionTopK::leader_scales is documented as "squared leader norms for L2" (and cosine uses unsquared norms), but this test helper feeds unsquared values for the L2 case. That makes the test data inconsistent with the public contract and can mask mistakes in distance computation. Consider squaring the L2 norms here so the tests exercise the intended inputs.

    let leader_scales = match metric {
        Metric::L2 => (0..leaders).map(|leader| (leader + 1) as f32).collect(),
        Metric::Cosine => (0..leaders)

diskann-pipnn/src/partition_kernel.rs:252

• For the L2 path, the SIMD chunk uses mul_add_simd (fused multiply-add) but the scalar tail uses norm - 2.0 * dot (non-fused). This can introduce small rounding differences between SIMD and tail elements, which can change ordering/tie behavior right at SIMD-width boundaries. Use f32::mul_add for the scalar tail so both paths compute the same value shape.

                |dot, norm| F::splat(arch, -2.0).mul_add_simd(dot, norm),
                |dot, norm| norm - 2.0 * dot,

[weiyaoluo (SeliMeli)]

weiyaoluo (@SeliMeli) please read the following Contributor License Agreement(CLA). If you agree with the CLA, please reply with the following information.

@microsoft-github-policy-service agree [company="{your company}"]

Options:

• (default - no company specified) I have sole ownership of intellectual property rights to my Submissions and I am not making Submissions in the course of work for my employer.

@microsoft-github-policy-service agree

• (when company given) I am making Submissions in the course of work for my employer (or my employer has intellectual property rights in my Submissions by contract or applicable law). I have permission from my employer to make Submissions and enter into this Agreement on behalf of my employer. By signing below, the defined term “You” includes me and my employer.

@microsoft-github-policy-service agree company="Microsoft"

Contributor License Agreement

@microsoft-github-policy-service agree company="Microsoft"

[Mark Hildebrand (hildebrandmw)]

I was hoping to get through the review today, but have run out of time. Please see the inline comments for the partial review.

One thing from a developer quality-of-life standpoint is that this PR makes heavy use of rstest to stamp out table-driven tests. As is, it increases the total number of test function in diskann from 354 to 1605! This adds tremendous noise to the test output since just about 78% of the tests are PiPNN related and the test names go on forever. E.g.

PASS [   0.014s] diskann graph::pipnn::partition_metric::tests::point_to_leader_scores_match_scalar_distances::case_4_inner_product::point_count_2_3::leader_count_1_1::dimensions_06_15

Many of these can be easily bundled up into an internal table driven test and would greatly increase the signal to noise ratio.

There is another orthogonal concern to using rstest this way and that is that Miri + nextest has a start up time that is proportional to the number of tests. You can try this out with diskann-wide where Miri takes forever to start up due to the 1000s of tests. Test spamming like this makes the Miri experience for developers unnecessarily worse.

Please consolidate tests that are merely using rstest to run through combinations of values.

[weiyaoluo (SeliMeli)]

Pushed b0ccf73 and cc6da38, and merged them into #1290, #1291, #1294, and #1295.

• Top-k: each kernel calls one function: select_top_k_ids (partition rows) or select_top_k_symmetric (leaf pair scan). The module docs describe both functions and where they are used. The fixed-size nearest sets stay internal (partition k = 1, 2, 3, 8, 10; leaf k = 1, 2, 3), and the AVX2 measurements that justify them are in comments next to the code.
• Unsafe: one SIMD load (load_group), bounded by &[f32; LANES]. The nightly Miri step runs two tests that reach all three call sites.
• Shapes: the kernels share check_output_rows and a checked distance_scratch. Top-k checks its shape preconditions only with debug assertions.
• Tests: the kernel happy paths run all four metrics across SIMD group boundaries on every architecture the test machine supports. The #[values] cross products are gone; at this layer, PiPNN has 111 of the 435 diskann lib tests.

[Mark Hildebrand (hildebrandmw)]

Thanks - this has been quite a journey! A few small last comments on my end mostly about protecting ourself against future changes of the SIMD width assumption.

Other than that, happy to see this merge!