Better decomp for torch.eye by jansel · Pull Request #163386 · pytorch/pytorch

jansel · 2025-09-20T00:00:50Z

Stack from ghstack (oldest at bottom):

[ghstack-poisoned]

pytorch-bot · 2025-09-20T00:00:53Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/163386

📄 Preview Python docs built from this PR
📄 Preview C++ docs built from this PR
❓ Need help or want to give feedback on the CI? Visit the bot commands wiki

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

✅ No Failures

As of commit a869842 with merge base 51152ef ():
💚 Looks good so far! There are no failures yet. 💚

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

torch/_refs/__init__.py

[ghstack-poisoned]

vadimkantorov · 2025-09-20T00:54:50Z

btw does uint8/uint16 also work here? or no impls for unsigned?

jansel · 2025-09-20T04:00:45Z

Not sure if all backends implement those types.

Skylion007 · 2025-09-20T17:07:02Z

Not sure if all backends implement those types.

More reason to factor this out in their own utility. It would be good to have a robust check here we could grab the min dtype that is supported on target device.

Skylion007

This is subtly complicated to do, and we should factor it out into its own function.

torch/_refs/__init__.py

[ghstack-poisoned]

jansel · 2025-09-21T05:41:15Z

The dtype thing wasn't the main reason for this change, it matters close to zero for performance. Just an incidental change. (These will be registers in a memory-bound kernel.)

The real improvement here was removing the torch.where call from the common path which was causing issues for complex numbers.

[ghstack-poisoned]

eellison · 2025-09-22T12:50:06Z

torch/_refs/__init__.py


+def _strength_reduce_integer(val: int) -> torch.dtype:
+    for possible_dtype in (torch.uint8, torch.uint16, torch.int32):
+        if val <= torch.iinfo(possible_dtype).max:


nit: fix for unbacked. If adding guards here is intentional: guard_or_False. Otherwise: statically_known_true(val <= torch.iinfo(possible_dtype).max)

This is only called for int, for SymInt we use int64.

pytorchmergebot · 2025-09-22T16:11:47Z

Starting merge as part of PR stack under #163434

pytorchmergebot · 2025-09-22T21:46:47Z

Starting merge as part of PR stack under #163415

Fixes #163372 Pull Request resolved: #163415 Approved by: https://github.com/Skylion007 ghstack dependencies: #163386, #163398, #163387, #163414

Fixes #163082 Pull Request resolved: #163419 Approved by: https://github.com/Skylion007, https://github.com/mlazos ghstack dependencies: #163386, #163398, #163387, #163414, #163415

Fixes #163300 Pull Request resolved: #163434 Approved by: https://github.com/drisspg ghstack dependencies: #163386, #163398, #163387, #163414, #163415, #163419

Fixes #163275 Pull Request resolved: #163393 Approved by: https://github.com/eellison, https://github.com/coconutruben ghstack dependencies: #163386, #163398, #163387, #163414, #163415, #163419, #163434

Fixes #162624 Fixes #162586 Pull Request resolved: #163412 Approved by: https://github.com/eellison ghstack dependencies: #163386, #163398, #163387, #163414, #163415, #163419, #163434, #163393

Fixes #163257 Pull Request resolved: #163422 Approved by: https://github.com/eellison ghstack dependencies: #163386, #163398, #163387, #163414, #163415, #163419, #163434, #163393, #163412

Fixes #163420 Pull Request resolved: #163481 Approved by: https://github.com/eellison ghstack dependencies: #163386, #163398, #163387, #163414, #163415, #163419, #163434, #163393, #163412, #163422

Fixes #163449 Pull Request resolved: #163520 Approved by: https://github.com/eellison ghstack dependencies: #163386, #163398, #163387, #163414, #163415, #163419, #163434, #163393, #163412, #163422, #163481

Fixes #163457 Pull Request resolved: #163482 Approved by: https://github.com/eellison ghstack dependencies: #163386, #163398, #163387, #163414, #163415, #163419, #163434, #163393, #163412, #163422, #163481, #163520

This reverts commit a8cd437. See #163481 (comment) This PR might also cause issues with cudagraphs. Pull Request resolved: #163737 Approved by: https://github.com/ezyang ghstack dependencies: #163386, #163398, #163387, #163414, #163415, #163419, #163434, #163393, #163412, #163422, #163481, #163520, #163482

Pull Request resolved: pytorch#163386 Approved by: https://github.com/eellison

Fixes pytorch#163328 Pull Request resolved: pytorch#163398 Approved by: https://github.com/eellison ghstack dependencies: pytorch#163386

Fixes pytorch#163243 Fixes pytorch#162561 Pull Request resolved: pytorch#163387 Approved by: https://github.com/eellison ghstack dependencies: pytorch#163386, pytorch#163398

Fixes pytorch#161484 Pull Request resolved: pytorch#163414 Approved by: https://github.com/Skylion007 ghstack dependencies: pytorch#163386, pytorch#163398, pytorch#163387

…163415) Fixes pytorch#163372 Pull Request resolved: pytorch#163415 Approved by: https://github.com/Skylion007 ghstack dependencies: pytorch#163386, pytorch#163398, pytorch#163387, pytorch#163414

Fixes pytorch#163082 Pull Request resolved: pytorch#163419 Approved by: https://github.com/Skylion007, https://github.com/mlazos ghstack dependencies: pytorch#163386, pytorch#163398, pytorch#163387, pytorch#163414, pytorch#163415

Fixes pytorch#163300 Pull Request resolved: pytorch#163434 Approved by: https://github.com/drisspg ghstack dependencies: pytorch#163386, pytorch#163398, pytorch#163387, pytorch#163414, pytorch#163415, pytorch#163419

Fixes pytorch#163275 Pull Request resolved: pytorch#163393 Approved by: https://github.com/eellison, https://github.com/coconutruben ghstack dependencies: pytorch#163386, pytorch#163398, pytorch#163387, pytorch#163414, pytorch#163415, pytorch#163419, pytorch#163434

…63412) Fixes pytorch#162624 Fixes pytorch#162586 Pull Request resolved: pytorch#163412 Approved by: https://github.com/eellison ghstack dependencies: pytorch#163386, pytorch#163398, pytorch#163387, pytorch#163414, pytorch#163415, pytorch#163419, pytorch#163434, pytorch#163393

Fixes pytorch#163257 Pull Request resolved: pytorch#163422 Approved by: https://github.com/eellison ghstack dependencies: pytorch#163386, pytorch#163398, pytorch#163387, pytorch#163414, pytorch#163415, pytorch#163419, pytorch#163434, pytorch#163393, pytorch#163412

Fixes pytorch#163420 Pull Request resolved: pytorch#163481 Approved by: https://github.com/eellison ghstack dependencies: pytorch#163386, pytorch#163398, pytorch#163387, pytorch#163414, pytorch#163415, pytorch#163419, pytorch#163434, pytorch#163393, pytorch#163412, pytorch#163422

Fixes pytorch#163449 Pull Request resolved: pytorch#163520 Approved by: https://github.com/eellison ghstack dependencies: pytorch#163386, pytorch#163398, pytorch#163387, pytorch#163414, pytorch#163415, pytorch#163419, pytorch#163434, pytorch#163393, pytorch#163412, pytorch#163422, pytorch#163481

Fixes pytorch#163457 Pull Request resolved: pytorch#163482 Approved by: https://github.com/eellison ghstack dependencies: pytorch#163386, pytorch#163398, pytorch#163387, pytorch#163414, pytorch#163415, pytorch#163419, pytorch#163434, pytorch#163393, pytorch#163412, pytorch#163422, pytorch#163481, pytorch#163520

This reverts commit a8cd437. See pytorch#163481 (comment) This PR might also cause issues with cudagraphs. Pull Request resolved: pytorch#163737 Approved by: https://github.com/ezyang ghstack dependencies: pytorch#163386, pytorch#163398, pytorch#163387, pytorch#163414, pytorch#163415, pytorch#163419, pytorch#163434, pytorch#163393, pytorch#163412, pytorch#163422, pytorch#163481, pytorch#163520, pytorch#163482

Fixes #163420 Pull Request resolved: #163481 Approved by: https://github.com/eellison ghstack dependencies: #163386, #163398, #163387, #163414, #163415, #163419, #163434, #163393, #163412, #163422

Fixes #163449 Pull Request resolved: #163520 Approved by: https://github.com/eellison ghstack dependencies: #163386, #163398, #163387, #163414, #163415, #163419, #163434, #163393, #163412, #163422, #163481

Fixes #163457 Pull Request resolved: #163482 Approved by: https://github.com/eellison ghstack dependencies: #163386, #163398, #163387, #163414, #163415, #163419, #163434, #163393, #163412, #163422, #163481, #163520

This reverts commit a8cd437. See #163481 (comment) This PR might also cause issues with cudagraphs. Pull Request resolved: #163737 Approved by: https://github.com/ezyang ghstack dependencies: #163386, #163398, #163387, #163414, #163415, #163419, #163434, #163393, #163412, #163422, #163481, #163520, #163482

ghstack-source-id: 0e21cec Pull-Request: pytorch/pytorch#163386

Update

c42e752

[ghstack-poisoned]

This was referenced Sep 19, 2025

[dynamo] Fix issue with namedtuple slicing #163351

Closed

[inductor] Fallback on strided complex add #163387

Closed

Fix prims.broadcast_in_dim functionalization #163377

Closed

vadimkantorov reviewed Sep 20, 2025

View reviewed changes

torch/_refs/__init__.py Outdated Show resolved Hide resolved

jansel mentioned this pull request Sep 20, 2025

[inductor] Support out_dtype arg to matmul #163393

Closed

jansel added the topic: not user facing topic category label Sep 20, 2025

Update

3ee9d0b

[ghstack-poisoned]

jansel mentioned this pull request Sep 20, 2025

[inductor] Fix bug where viewed outputs get padded #163398

Closed

jansel requested a review from eellison September 20, 2025 15:23

This was referenced Sep 20, 2025

[dynamo] Fix TorchFunctionMode handling with get_rng_state #163412

Closed

[inductor] Skip test_baddmm on XPU #163414

Closed

[inductor] Don't require_dense for grid_sampler_2d_backward #163415

Closed

Skylion007 reviewed Sep 20, 2025

View reviewed changes

torch/_refs/__init__.py Show resolved Hide resolved

This was referenced Sep 20, 2025

[inductor] libdevice.sqrt => tl.sqrt_rn #163419

Closed

[inductor] Fix error from custom CUDA allocators #163422

Closed

[inductor] Freeze layouts in FlexAttention #163434

Closed

Update

40a4d4b

[ghstack-poisoned]

Update

a869842

[ghstack-poisoned]

This was referenced Sep 22, 2025

[inductor] Fix issue with scalar arg handling #163481

Closed

[inductor] Fix divmod error in decomp #163482

Closed

eellison approved these changes Sep 22, 2025

View reviewed changes

jansel mentioned this pull request Sep 22, 2025

[inductor] Fix bugs in emulate_precision_casts #163520

Closed

dsashidh pushed a commit to dsashidh/pytorch that referenced this pull request Sep 26, 2025

Better decomp for torch.eye (pytorch#163386)

4caefbb

Pull Request resolved: pytorch#163386 Approved by: https://github.com/eellison

github-actions bot deleted the gh/jansel/534/head branch October 23, 2025 02:13

Khanaksahu pushed a commit to Khanaksahu/pytorch that referenced this pull request Nov 17, 2025

Better decomp for torch.eye

7f95bbd

ghstack-source-id: 0e21cec Pull-Request: pytorch/pytorch#163386

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Better decomp for torch.eye#163386

Better decomp for torch.eye#163386
jansel wants to merge 4 commits intogh/jansel/534/basefrom
gh/jansel/534/head

jansel commented Sep 20, 2025 •

edited

Loading

Uh oh!

pytorch-bot bot commented Sep 20, 2025 •

edited

Loading

Uh oh!

Uh oh!

vadimkantorov commented Sep 20, 2025

Uh oh!

jansel commented Sep 20, 2025

Uh oh!

Skylion007 commented Sep 20, 2025

Uh oh!

Skylion007 left a comment

Uh oh!

Uh oh!

jansel commented Sep 21, 2025 •

edited

Loading

Uh oh!

eellison Sep 22, 2025

Uh oh!

jansel Sep 22, 2025

Uh oh!

pytorchmergebot commented Sep 22, 2025

Uh oh!

pytorchmergebot commented Sep 22, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

jansel commented Sep 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Sep 20, 2025 • 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/163386

✅ No Failures

Uh oh!

Uh oh!

vadimkantorov commented Sep 20, 2025

Uh oh!

jansel commented Sep 20, 2025

Uh oh!

Skylion007 commented Sep 20, 2025

Uh oh!

Skylion007 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

jansel commented Sep 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

eellison Sep 22, 2025

Choose a reason for hiding this comment

Uh oh!

jansel Sep 22, 2025

Choose a reason for hiding this comment

Uh oh!

pytorchmergebot commented Sep 22, 2025

Uh oh!

pytorchmergebot commented Sep 22, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

jansel commented Sep 20, 2025 •

edited

Loading

pytorch-bot bot commented Sep 20, 2025 •

edited

Loading

jansel commented Sep 21, 2025 •

edited

Loading