[dynamic shapes] prims_common non_overlapping_and_dense #160462

pianpwk · 2025-08-12T21:10:36Z

Differential Revision: D80120333

cc @ezyang @SherlockNoMad @EikanWang @jgong5 @wenzhe-nrv

pytorch-bot · 2025-08-12T21:10:40Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/160462

📄 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 or our office hours

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

✅ No Failures

As of commit 5670291 with merge base 16ada80 ():
💚 Looks good so far! There are no failures yet. 💚

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

facebook-github-bot · 2025-08-12T21:10:48Z

This pull request was exported from Phabricator. Differential Revision: D80120333

Summary: Rollback Plan: Differential Revision: D80120333

facebook-github-bot · 2025-08-13T00:00:34Z

This pull request was exported from Phabricator. Differential Revision: D80120333

Summary: Rollback Plan: Differential Revision: D80120333

facebook-github-bot · 2025-08-13T01:06:58Z

This pull request was exported from Phabricator. Differential Revision: D80120333

laithsakka · 2025-08-13T06:45:59Z

torch/_prims_common/__init__.py

            continue

-        if guard_size_oblivious(stride != expected_stride):
+        if not stride_eq(stride, expected_stride):


instead of defining this function stride_eq would it make sense to have this loop in a utility function that is also called from contiguous and use the same approach we use there. which is

def func check_contiguity(sizes, strides): expected_stride = 1 expected_stride_max = 1 for x, y in reversed(tuple(zip(sizes, strides))): # Skips checking strides when a dimension has length 1. if guard_or_false(x == 1): continue if guard_or_true(y != expected_stride) and guard_or_true( y != expected_stride_max ): return False # We symbolically check both paths to maximize the cases where this function # returns true. This is because make_contiguous_strides_for adds the max # symbolically, and in some other situations the max might not be there. # And we want to ensure we return true in both cases. expected_stride_max *= x if is_nested_int(x) else sym_max(x, 1) # type:ignore[assignment] expected_stride *= x return True

torch/_prims_common/__init__.py

laithsakka · 2025-08-13T06:52:59Z

torch/_prims_common/__init__.py



-def is_non_overlapping_and_dense(a: Tensor) -> bool:
+def _is_non_overlapping_and_dense(sizes, strides) -> bool:


call the function _is_non_overlapping_and_dense_or_false.
"""may return false instead of throwing DDE at data dependency when the result is uknown"

Summary: Rollback Plan: Differential Revision: D80120333

facebook-github-bot · 2025-08-13T22:09:10Z

This pull request was exported from Phabricator. Differential Revision: D80120333

Summary: Pull Request resolved: #160462 Test Plan: test_dynamic_shapes Rollback Plan: Differential Revision: D80120333

facebook-github-bot · 2025-08-15T18:41:16Z

This pull request was exported from Phabricator. Differential Revision: D80120333

laithsakka · 2025-08-15T18:48:58Z

test/test_dynamic_shapes.py


+    def test_prims_non_overlapping_and_dense(self):
+        shape_env = ShapeEnv()
+        cf = torch._prims_common.is_non_overlapping_and_dense


is_non_overlapping_and_dense-> is_non_overlapping_and_dense_or_false?

laithsakka · 2025-08-15T18:52:55Z

torch/_prims_common/__init__.py

-    for length, stride in lengths_and_strides:
-        if guard_or_false(length == 1):
-            continue
+def is_non_overlapping_and_dense(a: Tensor) -> bool:


is this for BC compatibility?
In my mind calling is_non_overlapping_and_dense should throw DDE
calling _is_non_overlapping_and_dense_or_false should not
now this is sort of alias for _is_non_overlapping_and_dense_or_false
meaning unbacked semantics are implicit to users calling is_non_overlapping_and_dense

maybe its ok

laithsakka · 2025-08-15T18:53:53Z

torch/_prims_common/__init__.py

+            # for backed symbols, this is practically a < operation
+            # for unbacked, we return True if < is statically known,
+            # then try to answer this symbolically, with stride ordering semantics
+            # (e.g. u0 < u0 is False, u0 < u1 is False with no axioms, u0 < 2 * u0 is True)


please add a comnent explaining why its ok if in the worst case for unbacked we picked a wrong order.

laithsakka · 2025-08-15T18:55:58Z

torch/_prims_common/__init__.py

-        def __ge__(self, other):
-            return guard_size_oblivious(self.stride >= other.stride)
+    # verify that sorted order was imposed (checks the "non-overlapping condition")
+    for i, j in zip(lengths_and_strides[:-1], lengths_and_strides[1:]):


why do you need this?

laithsakka · 2025-08-15T19:00:27Z

torch/_prims_common/__init__.py

-        def __ge__(self, other):
-            return guard_size_oblivious(self.stride >= other.stride)
+    # verify that sorted order was imposed (checks the "non-overlapping condition")
+    for i, j in zip(lengths_and_strides[:-1], lengths_and_strides[1:]):


does anything fail if you remove this? I do not see why we would need it?

Summary: Pull Request resolved: #160462 Test Plan: test_dynamic_shapes Rollback Plan: Reviewed By: laithsakka Differential Revision: D80120333

facebook-github-bot · 2025-08-15T19:32:13Z

This pull request was exported from Phabricator. Differential Revision: D80120333

Summary: Pull Request resolved: #160462 Test Plan: test_dynamic_shapes Rollback Plan: Reviewed By: laithsakka Differential Revision: D80120333

facebook-github-bot · 2025-08-15T20:58:53Z

This pull request was exported from Phabricator. Differential Revision: D80120333

Summary: Pull Request resolved: #160462 Test Plan: test_dynamic_shapes Rollback Plan: Reviewed By: laithsakka Differential Revision: D80120333

facebook-github-bot · 2025-08-15T22:48:34Z

This pull request was exported from Phabricator. Differential Revision: D80120333

Summary: Pull Request resolved: #160462 Test Plan: test_dynamic_shapes Rollback Plan: Reviewed By: laithsakka Differential Revision: D80120333

facebook-github-bot · 2025-08-18T16:18:54Z

This pull request was exported from Phabricator. Differential Revision: D80120333

Summary: Pull Request resolved: #160462 Test Plan: test_dynamic_shapes Rollback Plan: Reviewed By: laithsakka Differential Revision: D80120333

facebook-github-bot · 2025-08-18T21:41:35Z

This pull request was exported from Phabricator. Differential Revision: D80120333

facebook-github-bot · 2025-08-18T23:00:26Z

@pytorchbot merge -i

(Initiating merge automatically since Phabricator Diff has merged, merging with -i because oss signals were bypassed internally)

pytorchmergebot · 2025-08-18T23:02:16Z

Merge started

Your change will be merged while ignoring the following 1 checks: pull / linux-jammy-py3.13-clang12 / test (default, 5, 5, linux.4xlarge)

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here

Differential Revision: D80120333 Pull Request resolved: pytorch#160462 Approved by: https://github.com/laithsakka

pianpwk requested review from bobrenjc93 and laithsakka as code owners August 12, 2025 21:10

pytorch-bot bot added ciflow/inductor release notes: fx release notes category labels Aug 12, 2025

facebook-github-bot added the fx label Aug 12, 2025

facebook-github-bot added the fb-exported label Aug 12, 2025

pianpwk marked this pull request as draft August 12, 2025 21:10

facebook-github-bot pushed a commit that referenced this pull request Aug 13, 2025

[WIP][dynamic shapes] prims_common non_overlapping_and_dense (#160462)

3c407d9

Summary: Rollback Plan: Differential Revision: D80120333

facebook-github-bot force-pushed the export-D80120333 branch from 32b6307 to 3c407d9 Compare August 13, 2025 00:00

facebook-github-bot pushed a commit that referenced this pull request Aug 13, 2025

[WIP][dynamic shapes] prims_common non_overlapping_and_dense (#160462)

8e04bc0

Summary: Rollback Plan: Differential Revision: D80120333

facebook-github-bot force-pushed the export-D80120333 branch from 3c407d9 to 8e04bc0 Compare August 13, 2025 01:06

laithsakka reviewed Aug 13, 2025

View reviewed changes

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

laithsakka reviewed Aug 13, 2025

View reviewed changes

torch/_prims_common/__init__.py Show resolved Hide resolved

laithsakka reviewed Aug 13, 2025

View reviewed changes

facebook-github-bot pushed a commit that referenced this pull request Aug 13, 2025

[WIP][dynamic shapes] prims_common non_overlapping_and_dense (#160462)

8a8a691

Summary: Rollback Plan: Differential Revision: D80120333

facebook-github-bot force-pushed the export-D80120333 branch from 8e04bc0 to 8a8a691 Compare August 13, 2025 22:08

facebook-github-bot pushed a commit that referenced this pull request Aug 15, 2025

[dynamic shapes] prims_common non_overlapping_and_dense (#160462)

ae88f5a

Summary: Pull Request resolved: #160462 Test Plan: test_dynamic_shapes Rollback Plan: Differential Revision: D80120333

facebook-github-bot force-pushed the export-D80120333 branch from 8a8a691 to ae88f5a Compare August 15, 2025 18:41

pianpwk changed the title ~~[WIP][dynamic shapes] prims_common non_overlapping_and_dense~~ [dynamic shapes] prims_common non_overlapping_and_dense Aug 15, 2025

pianpwk requested a review from laithsakka August 15, 2025 18:41

pytorch-bot bot added the ciflow/trunk Trigger trunk jobs on your pull request label Aug 15, 2025

pianpwk marked this pull request as ready for review August 15, 2025 18:41

laithsakka approved these changes Aug 15, 2025

View reviewed changes

laithsakka reviewed Aug 15, 2025

View reviewed changes

facebook-github-bot force-pushed the export-D80120333 branch from ae88f5a to a1e2be4 Compare August 15, 2025 19:31

pianpwk added a commit that referenced this pull request Aug 15, 2025

[dynamic shapes] prims_common non_overlapping_and_dense (#160462)

0fa3ef6

Summary: Pull Request resolved: #160462 Test Plan: test_dynamic_shapes Rollback Plan: Reviewed By: laithsakka Differential Revision: D80120333

pianpwk force-pushed the export-D80120333 branch from a1e2be4 to 0fa3ef6 Compare August 15, 2025 20:51

pianpwk added a commit that referenced this pull request Aug 15, 2025

[dynamic shapes] prims_common non_overlapping_and_dense (#160462)

a9c5959

Summary: Pull Request resolved: #160462 Test Plan: test_dynamic_shapes Rollback Plan: Reviewed By: laithsakka Differential Revision: D80120333

pianpwk force-pushed the export-D80120333 branch from 0fa3ef6 to a9c5959 Compare August 15, 2025 20:58

facebook-github-bot force-pushed the export-D80120333 branch from a9c5959 to c2b5724 Compare August 15, 2025 22:48

facebook-github-bot force-pushed the export-D80120333 branch from c2b5724 to 8ff8bb0 Compare August 18, 2025 16:18

[dynamic shapes] prims_common non_overlapping_and_dense (#160462)

5670291

Summary: Pull Request resolved: #160462 Test Plan: test_dynamic_shapes Rollback Plan: Reviewed By: laithsakka Differential Revision: D80120333

pianpwk force-pushed the export-D80120333 branch from 8ff8bb0 to 5670291 Compare August 18, 2025 21:41

pytorchmergebot added the merging label Aug 18, 2025

pytorchmergebot added the Merged label Aug 19, 2025

pytorchmergebot closed this in 4cb3101 Aug 19, 2025

pytorchmergebot removed the merging label Aug 19, 2025

can-gaa-hou pushed a commit to can-gaa-hou/pytorch that referenced this pull request Aug 22, 2025

[dynamic shapes] prims_common non_overlapping_and_dense (pytorch#160462)

771dce0

Differential Revision: D80120333 Pull Request resolved: pytorch#160462 Approved by: https://github.com/laithsakka

markc-614 pushed a commit to markc-614/pytorch that referenced this pull request Sep 17, 2025

[dynamic shapes] prims_common non_overlapping_and_dense (pytorch#160462)

c68b1ed

Differential Revision: D80120333 Pull Request resolved: pytorch#160462 Approved by: https://github.com/laithsakka

github-actions bot deleted the export-D80120333 branch September 18, 2025 02:08



		def is_non_overlapping_and_dense(a: Tensor) -> bool:
		def _is_non_overlapping_and_dense(sizes, strides) -> bool:

[dynamic shapes] prims_common non_overlapping_and_dense #160462

[dynamic shapes] prims_common non_overlapping_and_dense #160462

Uh oh!

Conversation

pianpwk commented Aug 12, 2025 • edited by pytorch-bot bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Aug 12, 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/160462

✅ No Failures

Uh oh!

facebook-github-bot commented Aug 12, 2025

Uh oh!

facebook-github-bot commented Aug 13, 2025

Uh oh!

facebook-github-bot commented Aug 13, 2025

Uh oh!

laithsakka Aug 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

laithsakka Aug 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot commented Aug 13, 2025

Uh oh!

facebook-github-bot commented Aug 15, 2025

Uh oh!

laithsakka Aug 15, 2025

Choose a reason for hiding this comment

Uh oh!

laithsakka Aug 15, 2025

Choose a reason for hiding this comment

Uh oh!

laithsakka Aug 15, 2025

Choose a reason for hiding this comment

Uh oh!

laithsakka Aug 15, 2025

Choose a reason for hiding this comment

Uh oh!

laithsakka Aug 15, 2025

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot commented Aug 15, 2025

Uh oh!

facebook-github-bot commented Aug 15, 2025

Uh oh!

facebook-github-bot commented Aug 15, 2025

Uh oh!

facebook-github-bot commented Aug 18, 2025

Uh oh!

facebook-github-bot commented Aug 18, 2025

Uh oh!

facebook-github-bot commented Aug 18, 2025

Uh oh!

pytorchmergebot commented Aug 18, 2025

Merge started

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

pianpwk commented Aug 12, 2025 •

edited by pytorch-bot bot

Loading

pytorch-bot bot commented Aug 12, 2025 •

edited

Loading

laithsakka Aug 13, 2025 •

edited

Loading

laithsakka Aug 13, 2025 •

edited

Loading