Skip to content

[benchmark] Simply Inputs more.#1077

Merged
hildebrandmw merged 14 commits into
mainfrom
mhildebr/simplify-any
May 19, 2026
Merged

[benchmark] Simply Inputs more.#1077
hildebrandmw merged 14 commits into
mainfrom
mhildebr/simplify-any

Conversation

@hildebrandmw
Copy link
Copy Markdown
Contributor

@hildebrandmw hildebrandmw commented May 15, 2026

Move the Input trait in diskann-benchmark-runner over to a strongly typed representation and remove Any from the public API. This finishes the large work in streamlining the benchmark API. The new interface is

pub trait Input: Sized + std::fmt::Debug + 'static {
    type Raw: serde::de::DeserializeOwned + serde::Serialize;
    fn tag() -> &'static str;
    fn from_raw(raw: Self::Raw, checker: &mut Checker) -> anyhow::Result<Self>;
    fn serialize(&self) -> anyhow::Result<serde_json::Value>;
    fn example() -> Self::Raw;
}

with the big changes being the introduction of the Raw associated type, changing try_deserialize for from_raw, adding serialize (for custom serialization) and example returning Self::Raw instead of raw serde_json::Value. The idea behind Raw is that it is used to replace the CheckDeserialization trait and allows decoupling of an unvalidated raw deserialized struct from the final validated object.

This PR also removes the CheckDeserialization trait.

Suggested Reviewing Order

  • diskann-benchmark-runner:
    • src/input.rs: This is where the bulk of the actual changes are. The now removed custom Any type is moved as an internal implementation detail in this module and is no longer user facing.
    • src/files.rs: The file path validation for InputFile is moved from CheckDeserialization to an inherent resolve method.
    • src/test/dim.rs: A mechanical change where Raw = Self is used for the Input definitions.
    • src/test/typed.rs: TypeInput is tweaked to use a separate type as the raw input to ensure there is some coverage there.
    • src/any.rs: Removed.
    • The rest of the changes are mostly mechanical updates of import paths.
  • diskann-benchmark:
    • src/inputs.mod.rs: Update the as_input! macro to the new form. Now that we aren't relying on CheckDeserialization, I switched over benchmarks to using ad-hoc validate inherent methods. This is somewhat of a stopgap to preserve the existing checks that mutate in-place.
    • README.md: Update some out-of-date examples and descriptions.
    • The rest of the changes are moving calls to check_deserialization to validate.
  • diskann-benchmark-simd: Simple port to the new Input trait.

@codecov-commenter
Copy link
Copy Markdown

codecov-commenter commented May 15, 2026

Codecov Report

❌ Patch coverage is 77.23577% with 56 lines in your changes missing coverage. Please review.
✅ Project coverage is 89.45%. Comparing base (205aad7) to head (3b40ab6).

Files with missing lines Patch % Lines
diskann-benchmark/src/inputs/graph_index.rs 59.61% 21 Missing ⚠️
diskann-benchmark/src/inputs/disk.rs 0.00% 11 Missing ⚠️
diskann-benchmark-simd/src/lib.rs 43.75% 9 Missing ⚠️
diskann-benchmark-runner/src/registry.rs 83.33% 4 Missing ⚠️
diskann-benchmark-runner/src/input.rs 94.54% 3 Missing ⚠️
diskann-benchmark-runner/src/test/dim.rs 78.57% 3 Missing ⚠️
diskann-benchmark-runner/src/test/typed.rs 90.90% 2 Missing ⚠️
diskann-benchmark/src/inputs/mod.rs 80.00% 2 Missing ⚠️
diskann-benchmark-runner/src/app.rs 66.66% 1 Missing ⚠️
Additional details and impacted files

Impacted file tree graph

@@            Coverage Diff             @@
##             main    #1077      +/-   ##
==========================================
- Coverage   89.46%   89.45%   -0.01%     
==========================================
  Files         459      458       -1     
  Lines       85482    85398      -84     
==========================================
- Hits        76475    76397      -78     
+ Misses       9007     9001       -6     
Flag Coverage Δ
miri 89.45% <77.23%> (-0.01%) ⬇️
unittests 89.08% <77.23%> (-0.01%) ⬇️

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

Files with missing lines Coverage Δ
diskann-benchmark-runner/src/benchmark.rs 84.25% <100.00%> (ø)
diskann-benchmark-runner/src/checker.rs 70.54% <ø> (-2.12%) ⬇️
diskann-benchmark-runner/src/files.rs 100.00% <100.00%> (ø)
...iskann-benchmark-runner/src/internal/regression.rs 97.69% <ø> (-0.01%) ⬇️
diskann-benchmark-runner/src/jobs.rs 96.77% <100.00%> (-0.06%) ⬇️
diskann-benchmark-runner/src/result.rs 97.84% <100.00%> (ø)
diskann-benchmark/src/inputs/exhaustive.rs 26.83% <100.00%> (ø)
diskann-benchmark/src/inputs/filters.rs 67.74% <100.00%> (-3.69%) ⬇️
diskann-benchmark-runner/src/app.rs 83.97% <66.66%> (-0.28%) ⬇️
diskann-benchmark-runner/src/test/typed.rs 97.10% <90.90%> (+0.50%) ⬆️
... and 7 more

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

@hildebrandmw hildebrandmw marked this pull request as ready for review May 16, 2026 01:04
@hildebrandmw hildebrandmw requested review from a team and Copilot May 16, 2026 01:04
Copy link
Copy Markdown
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

This PR completes the benchmark API streamlining by moving diskann-benchmark-runner::Input to a strongly-typed model (Raw DTO + from_raw validation + serialize) and removing the public Any/CheckDeserialization APIs, then mechanically porting benchmarks and inputs to the new interface.

Changes:

  • Replaced the runner’s dynamic Any + CheckDeserialization flow with Input { type Raw, from_raw(..), serialize(..), example() } and made the type-erasure an internal detail.
  • Migrated benchmark input validation from CheckDeserialization::check_deserialization to ad-hoc inherent validate(..) methods and InputFile::resolve(..).
  • Updated examples/docs and ported diskann-benchmark + diskann-benchmark-simd + runner tests to the new trait.

Reviewed changes

Copilot reviewed 21 out of 21 changed files in this pull request and generated 2 comments.

Show a summary per file
File Description
diskann-benchmark/src/inputs/mod.rs Updates as_input! macro to implement new Input API with Raw = Self, calling validate and serialize.
diskann-benchmark/src/inputs/graph_index.rs Ports graph-index inputs from CheckDeserialization to inherent validate + InputFile::resolve.
diskann-benchmark/src/inputs/filters.rs Ports filter inputs to inherent validate + InputFile::resolve.
diskann-benchmark/src/inputs/exhaustive.rs Ports exhaustive inputs to inherent validate + InputFile::resolve and updates nested validation calls.
diskann-benchmark/src/inputs/disk.rs Ports disk-index inputs to inherent validate and replaces file checks with resolve.
diskann-benchmark/src/backend/disk_index/benchmarks.rs Updates tolerance input to new Input trait (Raw, from_raw, serialize, example).
diskann-benchmark/README.md Updates README examples to the new registration and Input API, plus benchmark description hook.
diskann-benchmark-simd/src/lib.rs Ports SIMD benchmark inputs/tolerances to the new Input trait.
diskann-benchmark-runner/src/test/typed.rs Adjusts typed test inputs to use separate Raw DTOs and from_raw validation behavior.
diskann-benchmark-runner/src/test/dim.rs Ports dim test inputs to new Input trait methods.
diskann-benchmark-runner/src/result.rs Updates tests to reflect updated TypeInput::new signature.
diskann-benchmark-runner/src/registry.rs Switches registry internals to input::internal type-erased layer; hides dispatch methods from public API.
diskann-benchmark-runner/src/lib.rs Removes public re-exports of Any and CheckDeserialization; keeps Checker and Input public.
diskann-benchmark-runner/src/jobs.rs Updates job parsing to produce internal type-erased inputs and drops Checker tag plumbing.
diskann-benchmark-runner/src/internal/regression.rs Migrates regression pipeline to internal Any and updated deserialization path.
diskann-benchmark-runner/src/input.rs Introduces new Input trait shape and moves the type-erasure (Any/DynInput/Wrapper) into input::internal.
diskann-benchmark-runner/src/files.rs Replaces CheckDeserialization with InputFile::resolve for path validation/resolution.
diskann-benchmark-runner/src/checker.rs Removes CheckDeserialization and tag-based Checker::any helper; keeps path/output resolution utilities.
diskann-benchmark-runner/src/benchmark.rs Updates internal benchmark dispatch to use input::internal::Any instead of public Any.
diskann-benchmark-runner/src/app.rs Updates job serialization/checkpoint setup to use new Any::serialize() returning anyhow::Result.
diskann-benchmark-runner/src/any.rs Deleted (public Any removed).
Comments suppressed due to low confidence (2)

diskann-benchmark/src/inputs/exhaustive.rs:405

  • Spelling typo in comment: "Chcck" should be "Check".
        // Chcck that provided data type is compatible with `f32`.
        f32::check_converting_load(self.data_type)?;

diskann-benchmark/src/inputs/exhaustive.rs:506

  • Spelling typo in comment: "Chcck" should be "Check".
        // Chcck that provided data type is compatible with `f32`.
        f32::check_converting_load(self.data_type)?;

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

Comment thread diskann-benchmark/src/inputs/exhaustive.rs Outdated
Comment thread diskann-benchmark/src/inputs/mod.rs Outdated
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Comment thread diskann-benchmark-runner/src/input.rs Outdated
Comment thread diskann-benchmark-runner/src/input.rs Outdated
@hildebrandmw hildebrandmw merged commit 12d1a15 into main May 19, 2026
23 checks passed
@hildebrandmw hildebrandmw deleted the mhildebr/simplify-any branch May 19, 2026 17:26
arkrishn94 added a commit that referenced this pull request May 28, 2026
# DiskANN v0.53.0 Release Notes

## Breaking Changes

An AI generated, human reviewed list of changes is summarized below.

### Paged search overhauled — channel-based API
([#1078](#1078))

`PagedSearchState` and its `'static`-bound pause/resume model have been
replaced with an async, channel-based interface. The recommended way to
drive paged search is now via a `tokio::sync::mpsc` channel, with the
searcher embedded in an otherwise-`'static` future. See the [rendered
RFC](https://github.com/microsoft/DiskANN/blob/main/rfcs/01078-paged-search.md)
for the new shape. Callers wired against `PagedSearchState` must migrate
to the channel API.

Users of paged search via `wrapped_async::DiskANNIndex` that know their
inner futures will never suspend can use the new
`wrapped_async::DiskANNIndex::paged_search_no_await`; this will
efficiently run paged searches with minimal synchronization overhead.

### `DiskANNIndex::flat_search` removed
([#1076](#1076))

`DiskANNIndex::flat_search` and the `IdIterator` trait have been removed
from the `diskann` crate. Equivalent functionality lives on the new
inherent method `DiskIndexSearcher::flat_search` in `diskann-disk`. This
unblocks the experimental directions in #1067 and #983.

```rust
// Before
diskann_index.flat_search(query, ...)?;

// After
disk_index_searcher.flat_search(query, ...).await?;
```

### `DiskIndexSearcher::flat_search` now batched
([#1097](#1097))

The new `DiskIndexSearcher::flat_search` uses the bulk `pq_distances`
path instead of one-vector-at-a-time `Accessor::build_query_computer` +
`evaluate_similarity`. Downstream behavior is equivalent but tighter
resource bounds apply.

### `centroid` removed from PQ interfaces
([#1010](#1010))

The dataset-centroid argument has been removed from `FixedChunkPQTable`
construction, `populate`, and most other PQ APIs. The shift only ever
worked for L2 distance and was silently ignored for inner-product /
cosine, so passing it was a footgun. When an L2 shift is required, fold
it into the PQ pivots instead (the library now does this internally).

```rust
// Before
let table = FixedChunkPQTable::new(.., centroid, ..);

// After — drop the centroid argument
let table = FixedChunkPQTable::new(.., ..);
```

### Flat search interface
([#983](#983))

A new `flat` module in `diskann` adds a provider-agnostic brute-force
search surface, mirroring the shape of graph search. Backends implement
a single trait, `DistancesUnordered<C>` (in `flat/strategy.rs`), which
fuses iteration and distance computation, allowing any backend
(in-memory, quantized, disk, remote) to plug into a shared algorithm.
See the [rendered
RFC](https://github.com/microsoft/DiskANN/blob/main/rfcs/00983-flat-search.md).
This is additive but is the new canonical surface — direct ad-hoc
flat-search call sites should migrate.

### `bf_tree` extracted into `diskann-bftree` crate
([#1020](#1020))

The bf_tree provider has been moved out of `diskann-providers`
(previously at
`diskann-providers/src/model/graph/provider/async_/bf_tree/`) into a new
standalone `diskann-bftree` crate. Along with the move:

- Switched from PQ to spherical quantization.
- Dropped dependencies on `DeletionCheck`, `AsDeletionCheck`, and
`RemoveDeletedIdsAndCopy`.
- Simplified generics.

Consumers must update their `Cargo.toml` to depend on `diskann-bftree`
and update import paths.

### `direct_distance_impl` and `inner_product_raw` re-exposed
([#1081](#1081))

`direct_distance_impl` (free function) and
`FixedChunkPQTable::inner_product_raw` are `pub` again after being
privatized in #1044. Restored to unblock a downstream user. Not breaking
in the typical direction — this restores previously available API
surface.

### MinMax `recompress` takes a grid-scale parameter
([#1109](#1109))

The MinMax `recompress` API now accepts a grid-scale parameter. 

## New Features

- SIMD-optimized L2-squared norm
([#1107](#1107))
- Significantly faster bitmap computation
([#1099](#1099))
- Large speedup on the bitmap construction path used by filtered search.
- LLVM IR bloat regression check in CI
([#1083](#1083))
- CI now flags regressions in generated LLVM IR size, helping catch
unintended monomorphization blow-ups.
- Recall computation fix for under-k groundtruth
([#1069](#1069))

## Merged PRs

* Revise README for DiskANN3 by @harsha-simhadri in
#1046
* [CI] Try to fix publishing step by @hildebrandmw in
#1057
* [benchmark] Remove `DispatchRule` by @hildebrandmw in
#1064
* [benchmark] Automatic Input Registration by @hildebrandmw in
#1066
* Remove centroid from most PQ interfaces by @hildebrandmw in
#1010
* [diskann/disk] Remove `flat_search` from `DiskANNIndex` by
@hildebrandmw in #1076
* macos build and miri check to nightly by @harsha-simhadri in
#1058
* [API] Make some methods public again by @hildebrandmw in
#1081
* [benchmark] Simply `Inputs` more by @hildebrandmw in
#1077
* Turn on stack protection for the diskann-garnet NuGet build by
@jackmoffitt in #1082
* Fix options for diskann-garnet nuget pipeline by @jackmoffitt in
#1091
* [CI] add LLVM IR bloat regression check by @arazumov in
#1083
* Bump openssl from 0.10.79 to 0.10.80 by @dependabot[bot] in
#1093
* [Disk CI benchmarks] Use 1ES.Pool=diskann-github by @arazumov in
#869
* Fix recall computation for fewer than k groundtruth results by
@magdalendobson in #1069
* bf_tree migration away from diskann-providers by @JordanMaples in
#1020
* [RFC/diskann] Overhaul paged search by @hildebrandmw in
#1078
* Remove unsafe code from compute_vec_l2sq by @arazumov in
#1094
* Remove direct accessor call in `diskann-garnet` by @hildebrandmw in
#1098
* Refactor `DiskIndexSearcher::flat_search` to use batching by
@hildebrandmw in #1097
* [flat index] Flat Search Interface by @arkrishn94 in
#983
* migrating multi-hop tests from diskann-providers to diskann by
@JordanMaples in #928
* Significantly speed up bitmap computation by @magdalendobson in
#1099
* `compute_vecs_l2sq`: Replace scalar L2 Squared norm with
SIMD-optimized FastL2NormSquared by @arazumov in
#1107
* [minmax] Add grid scaling to recompress API by @arkrishn94 in
#1109

**Full Changelog**:
v0.52.0...v0.53.0
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants