Fix non-dense TensorSpan handling in SequenceEqual, Fill*, Resize*, IndexOf*, and ConcatenateOnDimension by Copilot · Pull Request #124225 · dotnet/runtime

Copilot · 2026-02-10T15:16:10Z

Description

This PR fixes incorrect behavior for non-dense (strided) tensor spans where methods were operating over raw underlying storage rather than logical tensor elements.

In addition to the original SequenceEqual fix, this now includes follow-up correctness and consistency updates requested in review feedback.

Changes Made

Fixed SequenceEqual for non-dense spans to compare logical elements only.
Fixed non-dense behavior in:
- FillGaussianNormalDistribution
- FillUniformDistribution
- Resize
- ResizeTo
- IndexOfMax
- IndexOfMin
- IndexOfMaxMagnitude
- IndexOfMinMagnitude
Updated dense fast paths to use FlattenedLength instead of _shape.LinearLength.
Ensured one-line control-flow bodies in modified code use braces.
Updated non-dense comparison path in SequenceEqual to align with IEquatable<T>.Equals semantics used by span-based comparisons.
Updated IndexOf*Fallback implementations to use ArrayPool<T>.Shared instead of allocating new arrays per call.
Refactored duplicated fill-test offset computation into shared ComputeLogicalOffsets.
Fixed ConcatenateOnDimension behavior for non-dense destination spans by handling destination layout correctly.
Added regression coverage for non-dense destination concatenation:
- TensorConcatenateNonDenseDestinationTests

Testing

Baseline build:
- ./build.sh clr+libs -rc release
Targeted tests:
- dotnet build /t:test src/libraries/System.Numerics.Tensors/tests/System.Numerics.Tensors.Tests.csproj
Result:
- System.Numerics.Tensors.Tests total: 5401
- Errors: 0
- Failed: 0

Original prompt

This section details on the original issue you should resolve

<issue_title>Clarify expected behavior of Tensor.SequenceEqual with non-dense spans</issue_title>
<issue_description>Given a non-dense ReadOnlyTensorSpan<T>, SequenceEqual will be false if the unobserved elements are not the same.

For example:
using System;
using System.Numerics.Tensors;

int[] d1 = [10, 20, 99, 99, 30, 40, 99, 99];
int[] d2 = [10, 20, 77, 77, 30, 40, 77, 77];

var ts1 = new ReadOnlyTensorSpan<int>(d1, [2, 2], [4, 1]);
var ts2 = new ReadOnlyTensorSpan<int>(d2, [2, 2], [4, 1]);

DumpTensorSpan(ts1);
DumpTensorSpan(ts2);

bool result = ts1.SequenceEqual(ts2);
Console.WriteLine($"SequenceEqual: {result}");

static void DumpTensorSpan<T>(ReadOnlyTensorSpan<T> span)
{
    Console.WriteLine($"Dense: {span.IsDense}");
    
    foreach (T x in span)
    {
        Console.Write($"{x}, ");
    }
    
    Console.WriteLine();
}
This will print
Dense: False
10, 20, 30, 40, 
Dense: False
10, 20, 30, 40, 
SequenceEqual: False
I am not sure if this is "by design" or not. If it is, the documentation may want to make a note of how non-dense spans behavior.

The unit tests for this appear to be commented out:

runtime/src/libraries/System.Numerics.Tensors/tests/TensorTests.cs

Line 644 in 591b3de

//public static void TensorSequenceEqualTests()

</issue_description>

Comments on the Issue (you are @copilot in this section)

Fixes Clarify expected behavior of Tensor.SequenceEqual with non-dense spans #124112

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

…or spans Co-authored-by: stephentoub <[email protected]>

dotnet-policy-service · 2026-02-10T16:22:03Z

Tagging subscribers to this area: @dotnet/area-system-numerics-tensors
See info in area-owners.md if you want to be subscribed.

Copilot

Pull request overview

Fixes Tensor.SequenceEqual / TensorSpan.SequenceEqual so that comparisons involving non-dense tensor spans compare only the logical elements (ignoring unobserved gap/padding in the underlying buffer), while preserving the dense fast-path.

Changes:

Updated ReadOnlyTensorSpan<T>.SequenceEqual to use a dense fast-path and a non-dense element-by-element slow-path.
Simplified the TensorSpan<T> overload to delegate to the ReadOnlyTensorSpan<T> implementation.
Added unit tests covering non-dense vs non-dense, dense vs non-dense (both directions), differing logical elements, and the TensorSpan<T> overload.

Reviewed changes

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

File	Description
src/libraries/System.Numerics.Tensors/src/System/Numerics/Tensors/netcore/Tensor.cs	Fixes `SequenceEqual` semantics for non-dense spans and centralizes logic via the `ReadOnlyTensorSpan<T>` overload.
src/libraries/System.Numerics.Tensors/tests/TensorTests.cs	Adds regression tests validating correct `SequenceEqual` behavior across dense/non-dense scenarios.

src/libraries/System.Numerics.Tensors/src/System/Numerics/Tensors/netcore/Tensor.cs

stephentoub · 2026-02-10T16:34:47Z

@copilot do other methods have the same or similar problem? e.g. IndexOf*, ResizeTo, Fill*, etc.? Please investigate and fix any other similar problems in the whole library.

…dd tests Co-authored-by: stephentoub <[email protected]>

Copilot · 2026-02-10T17:10:46Z

@copilot do other methods have the same or similar problem? e.g. IndexOf*, ResizeTo, Fill*, etc.? Please investigate and fix any other similar problems in the whole library.

Investigated all methods in the library that use MemoryMarshal.CreateSpan/CreateReadOnlySpan with LinearLength without density checks. Found and fixed the following in commit 99ba364:

FillGaussianNormalDistribution / FillUniformDistribution — wrote to raw buffer including gaps; now iterates logical elements via enumerator for non-dense spans
Resize (Tensor overload) — read from raw buffer of non-dense sliced tensors; now enumerates logical elements
ResizeTo (ReadOnlyTensorSpan overload) — read/wrote raw buffer for both source and destination; now uses element-by-element copy for non-dense spans
IndexOfMax / IndexOfMin / IndexOfMaxMagnitude / IndexOfMinMagnitude — searched raw buffer including gaps; now flattens to temp array for non-dense spans

Methods using TensorOperation.Invoke (like Sum, Product, Average, Equals, element-wise ops) are already correct — they iterate via AdjustToNextIndex. Also noted ConcatenateOnDimension has a similar issue with non-dense destinations but that requires a larger rewrite; proposed as follow-up work.

src/libraries/System.Numerics.Tensors/src/System/Numerics/Tensors/netcore/Tensor.cs

src/libraries/System.Numerics.Tensors/tests/TensorTests.cs

Copilot

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

Co-authored-by: stephentoub <[email protected]>

Copilot · 2026-03-05T21:41:20Z

@copilot try again

Implemented the follow-up fixes and retried with fresh validation in ec7f3ab. I added a new regression test for ConcatenateOnDimension with a non-dense destination span (the remaining gap from prior review feedback), then reran ./build.sh clr+libs -rc release and dotnet build /t:test src/libraries/System.Numerics.Tensors/tests/System.Numerics.Tensors.Tests.csproj (5401 tests, 0 failures). I also reran the code-review pass on the updated head and it reported no review comments.

tannergooding · 2026-03-17T15:51:11Z

src/libraries/System.Numerics.Tensors/src/System/Numerics/Tensors/netcore/Tensor.cs

+                // For non-dense destinations, concatenate into a temporary dense buffer,
+                // then copy element-by-element to respect the destination's stride layout.
+                T[] tempBuffer = ArrayPool<T>.Shared.Rent((int)totalLength);
+                try
+                {
+                    Span<T> tempSpan = tempBuffer.AsSpan(0, (int)totalLength);
+                    ConcatenateOnDimensionToSpan(dimension, tensors, destination, tempSpan);
+                    ReadOnlyTensorSpan<T> tempTensor = new ReadOnlyTensorSpan<T>(tempBuffer, 0, destination.Lengths, []);
+                    TensorOperation.Invoke<TensorOperation.CopyTo<T>, T, T>(tempTensor, destination);
+                }
+                finally
+                {
+                    ArrayPool<T>.Shared.Return(tempBuffer);
+                }


nit: Getting it working correctly is most important, so I think the temp buffer for right now is ok, but we should have a tracking issue to avoid the temporary buffer since it can add significant expense (due to amount of memory accesses/copying)

tannergooding · 2026-03-17T15:53:50Z

src/libraries/System.Numerics.Tensors/src/System/Numerics/Tensors/netcore/Tensor.cs

            random ??= Random.Shared;

-            for (int i = 0; i < span.Length; i++)
+            if (destination.IsDense)


Similar here. In general this is meant to be handled via the TensorDimensionSpan iterator, which allows getting the contiguous sequences more easily -or- via the TensorShape.AdjustToNextIndex APIs for more complex cases.

stephentoub · 2026-03-17T16:04:45Z

/ba-g deadletter

Initial plan

0d2cb5e

Copilot AI assigned Copilot and stephentoub Feb 10, 2026

Copilot started work on behalf of stephentoub February 10, 2026 15:16 View session

Fix SequenceEqual to compare only logical elements for non-dense tens…

3bc90c8

…or spans Co-authored-by: stephentoub <[email protected]>

github-actions bot added the needs-area-label An area label is needed to ensure this gets routed to the appropriate area owners label Feb 10, 2026

Copilot AI changed the title ~~[WIP] Clarify expected behavior of Tensor.SequenceEqual~~ Fix SequenceEqual for non-dense TensorSpan comparisons Feb 10, 2026

Copilot AI requested a review from stephentoub February 10, 2026 15:48

Copilot finished work on behalf of stephentoub February 10, 2026 15:48

stephentoub approved these changes Feb 10, 2026

View reviewed changes

stephentoub marked this pull request as ready for review February 10, 2026 16:20

Copilot AI review requested due to automatic review settings February 10, 2026 16:20

stephentoub added area-System.Numerics.Tensors and removed needs-area-label An area label is needed to ensure this gets routed to the appropriate area owners labels Feb 10, 2026

stephentoub requested a review from tannergooding February 10, 2026 16:21

Copilot started reviewing on behalf of stephentoub February 10, 2026 16:21 View session

Copilot AI reviewed Feb 10, 2026

View reviewed changes

src/libraries/System.Numerics.Tensors/src/System/Numerics/Tensors/netcore/Tensor.cs Show resolved Hide resolved

src/libraries/System.Numerics.Tensors/src/System/Numerics/Tensors/netcore/Tensor.cs Outdated Show resolved Hide resolved

Copilot started work on behalf of stephentoub February 10, 2026 16:35 View session

Fix non-dense span handling in Fill*, Resize*, IndexOf* methods and a…

99ba364

…dd tests Co-authored-by: stephentoub <[email protected]>

Copilot AI changed the title ~~Fix SequenceEqual for non-dense TensorSpan comparisons~~ Fix non-dense TensorSpan handling in SequenceEqual, Fill*, Resize*, and IndexOf* methods Feb 10, 2026

Copilot finished work on behalf of stephentoub February 10, 2026 17:13

Copilot AI requested a review from stephentoub February 10, 2026 17:13

stephentoub requested a review from Copilot February 10, 2026 18:09

Copilot started reviewing on behalf of stephentoub February 10, 2026 18:10 View session

stephentoub reviewed Feb 10, 2026

View reviewed changes

src/libraries/System.Numerics.Tensors/src/System/Numerics/Tensors/netcore/Tensor.cs Show resolved Hide resolved

src/libraries/System.Numerics.Tensors/tests/TensorTests.cs Show resolved Hide resolved

Copilot started work on behalf of stephentoub February 10, 2026 18:13 View session

Copilot AI reviewed Feb 10, 2026

View reviewed changes

Copilot started work on behalf of stephentoub March 5, 2026 21:16 View session

Add regression test for concatenate with non-dense destination span

ec7f3ab

Co-authored-by: stephentoub <[email protected]>

Copilot AI changed the title ~~Fix non-dense TensorSpan handling in SequenceEqual, Fill*, Resize*, and IndexOf* methods~~ Fix non-dense TensorSpan handling in SequenceEqual, Fill*, Resize*, IndexOf*, and ConcatenateOnDimension Mar 5, 2026

Copilot finished work on behalf of stephentoub March 5, 2026 21:42

Merge branch 'main' into copilot/clarify-sequence-equal-behavior

ddbbb2c

This was referenced Mar 11, 2026

[8.0] System.Exception: Failed to list devices - tvOS, iOS #125135

Open

XHarness device enumeration failure for tvos-device (exit code 71 GENERAL_FAILURE) #125371

Open

Merge branch 'main' into copilot/clarify-sequence-equal-behavior

1d382e3

Merge branch 'main' into copilot/clarify-sequence-equal-behavior

4ae60d1

build-analysis bot mentioned this pull request Mar 17, 2026

[wasm] OpenQA.Selenium.WebDriverTimeoutException: timeout: Timed out receiving message from renderer #117486

Open

tannergooding reviewed Mar 17, 2026

View reviewed changes

tannergooding approved these changes Mar 17, 2026

View reviewed changes

stephentoub merged commit c5c184a into main Mar 17, 2026
82 of 90 checks passed

stephentoub deleted the copilot/clarify-sequence-equal-behavior branch March 17, 2026 16:05

This was referenced Mar 17, 2026

System.Numerics.Tensors: Avoid temporary buffers in non-dense tensor operations (ConcatenateOnDimension, Fill*, IndexOf*) #125661

Closed

Optimize non-dense TensorSpan operations to avoid temporary buffer allocation #125663

Open

dotnet-maestro bot mentioned this pull request Mar 18, 2026

[main] Source code updates from dotnet/runtime dotnet/dotnet#5557

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix non-dense TensorSpan handling in SequenceEqual, Fill, Resize, IndexOf*, and ConcatenateOnDimension#124225

Fix non-dense TensorSpan handling in SequenceEqual, Fill, Resize, IndexOf*, and ConcatenateOnDimension#124225
stephentoub merged 10 commits intomainfrom
copilot/clarify-sequence-equal-behavior

Copilot AI commented Feb 10, 2026 •

edited

Loading

Uh oh!

dotnet-policy-service bot commented Feb 10, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

stephentoub commented Feb 10, 2026

Uh oh!

Copilot AI commented Feb 10, 2026

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI commented Mar 5, 2026

Uh oh!

tannergooding Mar 17, 2026

Uh oh!

tannergooding Mar 17, 2026

Uh oh!

stephentoub commented Mar 17, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

Copilot AI commented Feb 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Changes Made

Testing

Comments on the Issue (you are @copilot in this section)

Uh oh!

dotnet-policy-service bot commented Feb 10, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

stephentoub commented Feb 10, 2026

Uh oh!

Copilot AI commented Feb 10, 2026

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI commented Mar 5, 2026

Uh oh!

tannergooding Mar 17, 2026

Choose a reason for hiding this comment

Uh oh!

tannergooding Mar 17, 2026

Choose a reason for hiding this comment

Uh oh!

stephentoub commented Mar 17, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Copilot AI commented Feb 10, 2026 •

edited

Loading