镜像站点 · 本页由第三方 GitHub 只读镜像提供,非 GitHub 官方站点,不接受任何登录或凭据输入。前往 github.com
Skip to content

[benchmark-runner] Simplify Input representation - #1475

Merged
Mark Hildebrand (hildebrandmw) merged 3 commits into
mainfrom
mhildebr/simplify-input
Oct 14, 2026
Merged

Mark Hildebrand (hildebrandmw) merged 3 commits into
mainfrom
mhildebr/simplify-input

Conversation

@hildebrandmw

Copy link
Copy Markdown
Contributor

Consolidate input::Registered and input::internal::DynInput into just input::internal::Input. We can then use two constructors:

impl Input {
    // construct an `Input` for an actual input.
    fn new<T>() -> Self where T: Input;

    // construct a feature gated input.
    fn gated(tag: &'static str, features: Features) -> Self;
}

The trickiest update is Registry::register_input, which consolidates the two methods Registry::register_input<T: Input> and Registry::register_gated into a single method handling all registry conflicts. As a small bonus, this methods is no longer forced to be monomorphized for every input T, helping with compile times.

This technically breaks API with the removal of input::Registered<'_>, but our external and internal uses don't need this type.

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

🟡 Changes recommended

It removes the public Registry::input API without accounting for that additional compatibility break.

1 open finding
What changed in this PR

Simplifies benchmark input handling by consolidating dynamic input representations.

Changes:

  • Introduces a unified internal Input wrapper.
  • Consolidates input registration and conflict handling.
  • Updates tolerance and formatting paths to use the new representation.
File Description
app.rs Uses unified inputs for display and formatting.
benchmark.rs Returns owned tolerance inputs.
input.rs Implements the unified internal input representation.
jobs.rs Formats unified input references.
registry.rs Consolidates registration and stores unified inputs.

🧠 Review effort: Balanced


💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

@codecov-commenter

Copy link
Copy Markdown

Codecov Report

❌ Patch coverage is 95.00000% with 7 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.95%. Comparing base (177c03b) to head (b79e8bf).

Files with missing lines Patch % Lines
diskann-benchmark-runner/src/input.rs 89.39% 7 Missing ⚠️
Additional details and impacted files

Impacted file tree graph

@@           Coverage Diff           @@
##             main    #1475   +/-   ##
=======================================
  Coverage   90.95%   90.95%           
=======================================
  Files         586      586           
  Lines      118173   118173           
=======================================
+ Hits       107483   107488    +5     
+ Misses      10690    10685    -5     
Flag Coverage Δ
miri 90.95% <95.00%> (+<0.01%) ⬆️
unittests 90.65% <95.00%> (+<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/app.rs 86.80% <100.00%> (ø)
diskann-benchmark-runner/src/benchmark.rs 90.90% <100.00%> (ø)
diskann-benchmark-runner/src/jobs.rs 96.72% <100.00%> (ø)
diskann-benchmark-runner/src/registry.rs 93.62% <100.00%> (-0.13%) ⬇️
diskann-benchmark-runner/src/input.rs 92.36% <89.39%> (+4.12%) ⬆️

... and 5 files with indirect coverage changes

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

@hildebrandmw
Mark Hildebrand (hildebrandmw) merged commit ed8a2ec into main Oct 14, 2026
32 checks passed
@hildebrandmw
Mark Hildebrand (hildebrandmw) deleted the mhildebr/simplify-input branch October 14, 2026 21:41
Mark Hildebrand (hildebrandmw) added a commit that referenced this pull request Oct 26, 2026
# DiskANN v0.61.0

## Highlights

- **Quantization:** Added 3-bit spherical quantization with AVX2
acceleration, faster
  spherical encoding, and NEON float32 MinMax kernels.
- **In-memory indexing (experimental):** Added product quantization and
1-8-bit scalar
quantization, unified quantized representations, and AArch64
prefetching.
- **Multi-vector search:** Added `i8 x i8` and MinMax8 x MinMax4 MaxSim
kernels.
- **Fixes:** Preserve virtual start-point out-edges when writing disk
indexes, correctly
round-trip scalar quantization configurations without standard
deviation, and support
  zero-column mutable matrix iterators.

## Breaking changes and upgrading

Changes are concentrated in quantization APIs and experimental/tooling
crates.

* PQ training (#1467) - Training now returns `BasicTable`.
`SimplePivots` was removed.
Replace `flatten()` with `view_pivots().as_slice()` and use
`pivots_for(chunk)` for
  per-chunk access.
* PQ distance tables (#1458) - Change `ProcessInto` bounds to
`ProcessInto<f32>` and calls
from `process_into::<Metric>(...)` to `process_into::<Metric,
f32>(...)`.
* MaxSim kernels (#1429) - Generic kernel/erasure code now requires `T:
MaxSimElement` and
score buffers of `T::Score` (`f32` for floats, `i32` for `i8`). Update
`build_max_sim`
  error handling from `NotSupported` to `BuildMaxSimError`.
* `diskann-inmem` (#1485) - Replace `repr::Spherical` with
`repr::Quantized`. Wrap the
quantizer in `repr::quantized::spherical::Spherical::new(...)` when
configuring it.
  Configuration and `Rerank` types now live under `repr::quantized`.
* `diskann-benchmark-runner` (#1475) - `input::Registered` and public
`Registry::input` were
removed. Use the concrete `Input` implementation's `tag()` and
`example()` methods instead.

## All Changes
* Fix serialization round trip for QuantizationType::SQ without standard
deviation by @arrayka in #1461
* i8 x i8 MaxSim Kernels by @suri-kumkaran in
#1429
* Adding `subview_mut` to `MatrixMut` by @arkrishn94 in
#1463
* Add InsertResult::SuccessUpdate by @metajack in
#1449
* [diskann-quantization] PQ Infrastructure for `diskann-inmem` by
@hildebrandmw in #1458
* [diskann-quantization] Teach PQ training to return a `BasicTable`. by
@hildebrandmw in #1467
* [deps] Run `cargo-update` by @hildebrandmw in
#1452
* Support zero-column mutable matrix iterators by @arrayka in
#1462
* [diskann-quantization] Use `f32::NAN` instead of `core::f32::NAN`. by
@hildebrandmw in #1471
* [diskann-wide] Add `V4::new_checked_miri_uncached`. by @hildebrandmw
in #1480
* [diskann-wide] Add maddubs intrinsic wrappers to V3 and V4 by
@partychen in #1479
* Bump the github-actions group across 1 directory with 5 updates by
@dependabot[bot] in #1482
* [benchmark-runner] Simplify Input representation by @hildebrandmw in
#1475
* [diskann-quantization] Optimize spherical encoding scale search by
@xinyuwen2 in #1478
* [diskann-garnet] Implement XVCREATE, XVIMPORT by @metajack in
#1455
* [diskann-garnet] Support configurable start points and allow importing
start points by @metajack in
#1464
* [diskann-garnet] Add spherical 2-bit and 4-bit quantization by
@metajack in #1465
* [deps] Address depandabot issue 32 by @hildebrandmw in
#1496
* [diskann-disk] Preserve virtual start point out-edges when writing the
disk index by @doliawu in #1483
* [diskann-quantization] Add Neon F32 Minmax Kernels by @pfoxARM in
#1431
* [diskann-garnet] Add macOS arm64 support to NuGet publishing by
@metajack in #1495
* [diskann-disk] Require start points at index initialization. by
@hildebrandmw in #1497
* Expose metric-aware quantizer centroid distance by @cherryrhea in
#1395
* [diskann-inmem] Add support for Product Quantization by @hildebrandmw
in #1469
* Adds aarch64 prefetching into diskann-vector and diskann-inmem by
@pfoxARM in #1408
* [diskann-inmem] Add scalar quantization by @hildebrandmw in
#1484
* Bump the github-actions group with 3 updates by @dependabot[bot] in
#1505
* [diskann-inmem] Spherical Quantization Round 2 by @hildebrandmw in
#1485
* [diskann-quantization] Add SphericalQ 3-bit support and AVX2 kernel by
@xinyuwen2 in #1474
* [muilt-vetor] Add standalone MinMax8 x MinMax4 MaxSim kernel by
@partychen in #1494

## New Contributors
* @cherryrhea made their first contribution in
#1395

**Full Changelog**:
v0.60.0...v0.61.0

Co-authored-by: Mark Hildebrand <mhildebrand@microsoft.com>
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