WIP: Fix complex div broadcasting in CPU kernels by mshr-h · Pull Request #24409 · pytorch/executorch

mshr-h · 2026-11-02T01:53:09Z

Summary

The optimized and portable complex div kernels index both inputs using the output index. When an operand is broadcast, this can read past its buffer and produce incorrect results.

Use apply_binary_elementwise_fn to handle broadcasting, and require matching input/output dtypes before accessing typed data.

Add regression tests for complex64/128 covering scalar operands on either side, multidimensional broadcasting, empty tensors, and mismatched dtypes.

Test plan

Run the div tests in both CPU configurations:

cmake-out/kernels/test/optimized_kernels_test --gtest_filter='OpDiv*'
cmake-out/kernels/test/portable_kernels_test --gtest_filter='OpDiv*'

Result: 34 optimized CPU tests and 38 portable CPU tests passed.

pytorch-bot · 2026-11-02T01:53:13Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/24409

📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

⚠️ 27 Awaiting Approval

As of commit 135c5b6 with merge base b3b565b ():

AWAITING APPROVAL - The following workflows need approval before CI can run:

This comment was automatically generated by Dr. CI and updates every 15 minutes.

mshr-h · 2026-11-02T01:53:25Z

@pytorchbot label "release notes: ops & kernels"

Copilot

🟡 Changes recommended

The portable and optimized division helpers write broadcast results sequentially rather than using output strides, permuting results for channels-last tensors and requiring a stride-aware fix with regression coverage.

1 open issue

Broadcast helper misorders channels-last outputs · New

What changed in this PR

This PR addresses the complex division broadcasting bug from #24408 in ExecuTorch’s portable and optimized CPU kernels.

Changes:

Replace direct input indexing with the shared binary elementwise broadcasting helper.
Require matching complex input and output dtypes before typed data access.
Add complex64/128 regression tests for scalar and multidimensional broadcasting, empty tensors, and dtype mismatches.

File	Description
`kernels/test/op_div_test.cpp`	Adds complex broadcasting and dtype-rejection tests.
`kernels/portable/cpu/op_div.cpp`	Uses the broadcasting helper and validates complex dtypes.
`kernels/optimized/cpu/op_div.cpp`	Applies the same changes to optimized complex division.

_{Review effort: Balanced}

Copilot

🟢 Approval recommended

No active review issues are reported for the complex-division broadcasting and dtype-validation changes.

1 open issue

Broadcast helper misorders channels-last outputs

_{Review effort: Balanced}

Copilot

🟢 Approval recommended

No active review issues remain for the complex-division broadcasting changes.

1 open issue

Broadcast helper misorders channels-last outputs

_{Review effort: Balanced}

Copilot

🟢 Approval recommended

No active review issues are recorded for the complex-division broadcasting changes.

No open issues 🎉

1 resolved since last review

Broadcast helper misorders channels-last outputs

_{Review effort: Balanced}

Copilot

🟢 Approval recommended

No active review issues block the complex-division broadcasting changes.

No open issues 🎉

_{Review effort: Balanced}

Copilot

🟡 Changes recommended

The stride-ignoring fast path remains enabled for mixed-rank channels-last tensors, causing both division kernels to return incorrectly ordered results.

1 open issue

Use stride-aware iteration for mixed-rank channels-last division · New

_{Review effort: Balanced}

Copilot

🟡 Changes recommended

The shared helper unconditionally disables the no-broadcast fast path, adding unnecessary axis traversal for equal-shaped contiguous operands and losing sequential memory access for equal-shaped channels-last tensors.

1 open issue

Preserve linear iteration for matching broadcast shapes and layouts · New

1 resolved since last review

Use stride-aware iteration for mixed-rank channels-last division

_{Review effort: Balanced}

JakeStevens · 2026-11-12T14:31:11Z

@mshr-h is this ready for review? it is marked draft and WIP

mshr-h · 2026-11-12T15:02:32Z

@JakeStevens thanks for checking. it's not ready for review. i'll ping you when it's ready.

Use broadcast-aware indexing for complex div in optimized and portable CPU kernels, and require matching complex input and output dtypes. Respect output strides while preserving linear iteration for matching dense layouts. Cover scalar and channels-last broadcasting, mixed ranks, empty tensors, bounded output resizing, and invalid inputs. Refs pytorch#24408

Copilot

🟡 Changes recommended

The traversal change needs channels-last masked_scatter regressions to verify sequential source-to-mask placement in both traversal paths and insufficient-source failure.

2 open issues

_{Review effort: Balanced}

Consume source values in logical output order for channels-last tensors while preserving the default-layout fast path. Add regression tests for matching shapes, mixed ranks, mask broadcasting, and insufficient sources. Include masked_scatter tests in the CMake suite.

Copilot

🟡 Changes recommended

The new masked_scatter path can produce incorrect results for channels-last sources because it consumes source elements in physical rather than logical order.

2 open issues

2 resolved since last review

_{Review effort: Max}

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Copilot

🟡 Changes recommended

The new source-stride remapping lacks channels-last source coverage, leaving incorrect physical-order consumption undetected by the existing tests.

1 open issue

Add channels-last source coverage for masked scatter stride remapping · New

2 resolved since last review

_{Review effort: Max}

…rage for masked scatter stride remapping' Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Copilot

🔵 Needs a closer look

Human assessment is needed because an unpublished finding identifies a complex-scalar/channels-last compatibility gap in both division implementations.

No open issues 🎉

1 resolved since last review

Add channels-last source coverage for masked scatter stride remapping

_{Review effort: Max}

Copilot AI balanced review requested due to automatic review settings November 2, 2026 01:53

meta-cla Bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Nov 2, 2026

pytorch-bot Bot added the release notes: ops & kernels Changes to the opset and any new / changed kernel implementations label Nov 2, 2026

Copilot started reviewing on behalf of mshr-h November 2, 2026 01:53

View session

Copilot AI reviewed Nov 2, 2026

View reviewed changes

Comment thread kernels/portable/cpu/op_div.cpp Resolved

Copilot AI balanced review requested due to automatic review settings November 2, 2026 13:30

Copilot started reviewing on behalf of mshr-h November 2, 2026 13:31

View session

Copilot AI reviewed Nov 2, 2026

View reviewed changes

executorch-triage Bot added the community: contribution PRs coming from community (excluding hardware partners) label Nov 2, 2026

SofianElmotiem mentioned this pull request Nov 2, 2026

[CPU] Complex scalar broadcasting fails or returns incorrect results #24408

Open

mshr-h force-pushed the fix-complex-div-broadcasting branch from 614b52b to 3d24ca9 Compare November 2, 2026 22:01

Copilot AI balanced review requested due to automatic review settings November 2, 2026 22:01

Copilot started reviewing on behalf of mshr-h November 2, 2026 22:01

View session

Copilot AI reviewed Nov 2, 2026

View reviewed changes

mshr-h force-pushed the fix-complex-div-broadcasting branch from 3d24ca9 to 01f6068 Compare November 3, 2026 06:26

Copilot AI balanced review requested due to automatic review settings November 3, 2026 06:26

Copilot started reviewing on behalf of mshr-h November 3, 2026 06:27

View session

Copilot AI reviewed Nov 3, 2026

View reviewed changes

mshr-h force-pushed the fix-complex-div-broadcasting branch from 01f6068 to acadf9b Compare November 3, 2026 11:04

Copilot AI balanced review requested due to automatic review settings November 3, 2026 11:04

Copilot started reviewing on behalf of mshr-h November 3, 2026 11:04

View session

Copilot AI reviewed Nov 3, 2026

View reviewed changes

mshr-h force-pushed the fix-complex-div-broadcasting branch from acadf9b to 9241ec6 Compare November 3, 2026 13:53

Copilot AI balanced review requested due to automatic review settings November 3, 2026 13:53

Copilot started reviewing on behalf of mshr-h November 3, 2026 13:54

View session

Copilot AI reviewed Nov 3, 2026

View reviewed changes

Comment thread kernels/portable/cpu/util/broadcast_util.h Resolved

Copilot AI balanced review requested due to automatic review settings November 3, 2026 14:16

Copilot started reviewing on behalf of mshr-h November 3, 2026 14:17

View session

Copilot AI reviewed Nov 3, 2026

View reviewed changes

Comment thread kernels/portable/cpu/util/broadcast_util.h Resolved

mshr-h marked this pull request as draft November 3, 2026 14:27

mshr-h force-pushed the fix-complex-div-broadcasting branch from 050b241 to fb256ec Compare November 3, 2026 15:56

mshr-h changed the title ~~Fix complex div broadcasting in CPU kernels~~ WIP: Fix complex div broadcasting in CPU kernels Nov 3, 2026

mergennachin requested a review from JakeStevens November 11, 2026 15:18

mshr-h force-pushed the fix-complex-div-broadcasting branch from fb256ec to 21a37dd Compare November 15, 2026 04:48

mshr-h force-pushed the fix-complex-div-broadcasting branch from 21a37dd to 3e058da Compare November 15, 2026 04:54

mshr-h requested a balanced review from Copilot November 15, 2026 04:58

Copilot started reviewing on behalf of mshr-h November 15, 2026 04:58

View session

Copilot AI reviewed Nov 15, 2026

View reviewed changes

Comment thread kernels/portable/cpu/util/broadcast_util.h Resolved

mshr-h force-pushed the fix-complex-div-broadcasting branch from 3e058da to 06aa91e Compare November 15, 2026 08:11

mshr-h requested a max review from Copilot November 15, 2026 08:15

Copilot started reviewing on behalf of mshr-h November 15, 2026 08:16

View session

Copilot AI reviewed Nov 15, 2026

View reviewed changes

Comment thread kernels/portable/cpu/op_masked_scatter.cpp Resolved

Comment thread kernels/test/op_div_test.cpp Resolved

Apply batched suggestions from code review

58c7746

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

mshr-h requested a max review from Copilot November 15, 2026 08:35

Copilot started reviewing on behalf of mshr-h November 15, 2026 08:36

View session

Copilot AI reviewed Nov 15, 2026

View reviewed changes

Comment thread kernels/test/op_masked_scatter_test.cpp Resolved

Potential fix for pull request finding 'Add channels-last source cove…

135c5b6

…rage for masked scatter stride remapping' Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

mshr-h requested a max review from Copilot November 15, 2026 08:50

Copilot started reviewing on behalf of mshr-h November 15, 2026 08:51

View session

Copilot AI reviewed Nov 15, 2026

View reviewed changes

Conversation

mshr-h commented Nov 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test plan

Uh oh!

pytorch-bot Bot commented Nov 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/24409

⚠️ 27 Awaiting Approval

Uh oh!

mshr-h commented Nov 2, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟢 Approval recommended

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟢 Approval recommended

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟢 Approval recommended

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟢 Approval recommended

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Uh oh!

JakeStevens commented Nov 12, 2026

Uh oh!

mshr-h commented Nov 12, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🔵 Needs a closer look

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

mshr-h commented Nov 2, 2026 •

edited

Loading

pytorch-bot Bot commented Nov 2, 2026 •

edited

Loading