Migrate XPU UT files from Upstream PR by shangerxin · Pull Request #3593 · intel/torch-xpu-ops

shangerxin · 2026-05-10T11:39:37Z

Summary

Migrate 11 files

Upstream PR

pytorch/pytorch#169241

XPU Issues

#3356
#3393

Copilot

Pull request overview

Read skill file(s): .github/skills/xpu-ops-pr-review/SKILL.md.

This PR migrates a batch of upstream TorchDynamo/Inductor-related XPU unit test files into test/xpu/ and updates the XPU skip list to exclude these new tests from execution (at least for now), aligning the repo’s test tree with upstream structure and future enablement.

Changes:

Added migrated Dynamo/Inductor-related XPU test modules under test/xpu/dynamo/ plus a helper module under test/xpu/.
Updated test/xpu/skip_list_common.py to skip the newly added test files.
Introduced an async compile/process-pool tracking helper with queue timing metrics.

Reviewed changes

Copilot reviewed 7 out of 12 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
`test/xpu/tracked_process_pool_xpu.py`	Adds a `ProcessPoolExecutor` wrapper that tracks queue depth/timing and reports metrics at exit.
`test/xpu/skip_list_common.py`	Adds the newly migrated test files to the global skip list (skipping whole files).
`test/xpu/dynamo/test_trace_rules_xpu.py`	Migrated trace-rules tests for Dynamo rule maps and inlining behavior.
`test/xpu/dynamo/test_structured_trace_xpu.py`	Migrated structured-logging/trace validation tests for Dynamo/Inductor flows.
`test/xpu/dynamo/test_regional_inductor_xpu.py`	Migrated extensive regional-inductor tests (annotations, invoke_subgraph, serialization, partitioning).
`test/xpu/dynamo/test_deviceguard_xpu.py`	Migrated DeviceGuard tests for mocked and real device interfaces.
`test/xpu/dynamo/test_debug_utils_xpu.py`	Migrated Dynamo debug utilities tests (env var generation, override routing, parsers).
`test/xpu/dynamo/test_compiler_bisector_xpu.py`	Migrated Inductor compiler-bisector integration tests.
`test/xpu/dynamo/test_activation_checkpointing_xpu.py`	Migrated activation-checkpointing via tags/selective AC tests.
`test/xpu/dynamo/test_aot_autograd_cache_xpu.py`	Migrated AOTAutograd cache tests and related integration coverage.
`test/xpu/dynamo/test_export_xpu.py`	Migrated Dynamo export tests (graph generation, constraints, error cases).
`test/xpu/dynamo/test_subclasses_xpu.py`	Migrated tensor subclass/nested tensor Dynamo tests.

+try:
+    from .utils import create_dummy_module_and_function
+except ImportError:
+    from utils import create_dummy_module_and_function
+
+


+    stats = _queue_stats
+    if stats.pool_count == 0:
+        return
+
+    timing = stats.timing
+    timing.sort()
+
+    log.info("AsyncCompile Metrics:")
+    log.info("  Pools %s", stats.pool_count)
+    log.info(
+        "  Items %d enqueued / %d dequeued", stats.enqueue_count, stats.dequeue_count
+    )
+    log.info("  Max Queue Depth: %d", stats.max_queue_depth)


+@unittest.skipIf(not TEST_GPU, "No GPU available.")
+class TestGPUDeviceGuard(torch._dynamo.test_case.TestCase):
+    """
+    Unit tests for the DeviceGuard class using a GPU=Interface.


+        self.assertEqual(out.subsystem, "pre_grad_graph")
+        self.assertEqual(out.bisect_number, 1)
+
+    @skipIfXpu(msg="XPU doesn't support cudagrah")


shangerxin · 2026-05-14T08:13:01Z

@copilot fix the lint issue on this PR.

lint job log: https://github.com/intel/torch-xpu-ops/actions/runs/25848307428/job/75948554362?pr=3593

instructions:

review the lint failures log via the link above
fix all lint errors in the relevant source files
ensure both python lint and clang format checks pass

Copilot

Pull request overview

Copilot reviewed 4 out of 8 changed files in this pull request and generated 2 comments.

+device_type = (
+    acc.type if (acc := torch.accelerator.current_accelerator(True)) else "cpu"
+)


    "profiler/test_execution_trace_xpu.py": None,
    "profiler/test_profiler_xpu.py": None,
    "export/test_hop_xpu.py": None,
+    "tracked_process_pool_xpu.py": None,


shangerxin · 2026-05-18T07:37:25Z

@copilot fix the lint issue on this PR.

lint job log: https://github.com/intel/torch-xpu-ops/actions/runs/25848307428/job/75948554362?pr=3593

instructions:

review the lint failures log via the link above
fix all lint errors in the relevant source files
ensure both python lint and clang format checks pass

…erxin/torch-xpu-ops into erxin/migrate_dynamo_cases

Copilot

Pull request overview

Copilot reviewed 4 out of 8 changed files in this pull request and generated 2 comments.

+device_type = (
+    acc.type if (acc := torch.accelerator.current_accelerator(True)) else "cpu"
+)


    "profiler/test_execution_trace_xpu.py": None,
    "profiler/test_profiler_xpu.py": None,
    "export/test_hop_xpu.py": None,
+    "tracked_process_pool_xpu.py": None,


Copilot

Pull request overview

Copilot reviewed 5 out of 9 changed files in this pull request and generated 2 comments.

Comments suppressed due to low confidence (1)

test/xpu/tracked_process_pool_xpu.py:22

The module-level device_type defined here is computed but never used anywhere in this file. Additionally, torch.accelerator.current_accelerator(True) passes check=True, which will raise on environments without an accelerator and would cause this helper module to fail at import time. Since device_type is unused, this block should be removed.

device_type = (
    acc.type if (acc := torch.accelerator.current_accelerator(True)) else "cpu"
)

    "profiler/test_execution_trace_xpu.py": None,
    "profiler/test_profiler_xpu.py": None,
    "export/test_hop_xpu.py": None,
+    "tracked_process_pool_xpu.py": None,


+import torch._thread_safe_fork
+
+
+device_type = (
+    acc.type if (acc := torch.accelerator.current_accelerator(True)) else "cpu"
+)


Copilot

Pull request overview

Copilot reviewed 5 out of 9 changed files in this pull request and generated 3 comments.

Comments suppressed due to low confidence (1)

test/xpu/tracked_process_pool_xpu.py:21

The device_type variable defined here is never used anywhere in this file. The upstream tracked_process_pool.py does not include this block. Since this is a verbatim copy of the upstream helper and device_type serves no purpose here, this block (and the resulting dependency on torch.accelerator) appears to be unnecessary dead code that should be removed.

device_type = (
    acc.type if (acc := torch.accelerator.current_accelerator(True)) else "cpu"
)

+import torch._thread_safe_fork
+from typing_extensions import ParamSpec
+
+device_type = (
+    acc.type if (acc := torch.accelerator.current_accelerator(True)) else "cpu"
+)


+    'test/xpu/dynamo/test_*.py',
+    'test/xpu/dynamo/test_subclasses_xpu.py',
+    'test/xpu/dynamo/test_structured_trace_xpu.py',


+
+        self.assertParses()
+
+    @skipIf(HAS_XPU_AND_TRITON, "No backend type associated with device type xpu")


Copilot

Pull request overview

Copilot reviewed 5 out of 9 changed files in this pull request and generated 5 comments.

    "profiler/test_execution_trace_xpu.py": None,
    "profiler/test_profiler_xpu.py": None,
    "export/test_hop_xpu.py": None,
+    "tracked_process_pool_xpu.py": None,


+device_type = (
+    acc.type if (acc := torch.accelerator.current_accelerator(True)) else "cpu"
+)


+from torch.testing._internal.inductor_utils import HAS_XPU_AND_TRITON
+from torch.testing._internal.triton_utils import requires_gpu_and_triton
+
+device_type = acc.type if (acc := torch.accelerator.current_accelerator()) else "cpu"


+
+        self.assertParses()
+
+    @skipIf(HAS_XPU_AND_TRITON, "No backend type associated with device type xpu")


+    'test/xpu/dynamo/test_*.py',
+    'test/xpu/dynamo/test_subclasses_xpu.py',
+    'test/xpu/dynamo/test_structured_trace_xpu.py',


Copilot AI review requested due to automatic review settings May 10, 2026 11:39

Copilot AI reviewed May 10, 2026

View reviewed changes

Copilot AI mentioned this pull request May 11, 2026

Fix cudagraph tests blocked by feature gap in xpugraph #3613

Closed

shangerxin added disable_e2e Disable all e2e test jobs for the PR disable_distributed Disable distributed UT test jobs for the PR labels May 12, 2026

shangerxin added 2 commits May 14, 2026 15:35

Migrate xpu uts from upstream

cd36f2f

Rebase and fix merge issue

4c0658c

shangerxin force-pushed the erxin/migrate_dynamo_cases branch from 2d1e53b to 4c0658c Compare May 14, 2026 07:42

shangerxin changed the title ~~Migrate11 XPU UT files from Upstream PR~~ Migrate XPU UT files from Upstream PR May 14, 2026

shangerxin requested a review from Copilot May 14, 2026 08:15

Copilot started reviewing on behalf of shangerxin May 14, 2026 08:16 View session

Copilot AI reviewed May 14, 2026

View reviewed changes

shangerxin added 3 commits May 19, 2026 14:14

Merge branch 'main' into erxin/migrate_dynamo_cases

6935876

Fix lint issue

17de33f

Merge branch 'erxin/migrate_dynamo_cases' of https://github.com/shang…

8ba84c7

…erxin/torch-xpu-ops into erxin/migrate_dynamo_cases

Copilot AI review requested due to automatic review settings May 19, 2026 06:15

Copilot started reviewing on behalf of shangerxin May 19, 2026 06:16 View session

Copilot AI reviewed May 19, 2026

View reviewed changes

shangerxin added 3 commits May 19, 2026 14:45

Ignore lint check for the test files comming from upstream pytorch repo

8ca8ef9

Ignore the lint error

85fef8d

remove the orig files generated by beyond compare 4

106c802

Copilot AI review requested due to automatic review settings May 19, 2026 07:12

Copilot started reviewing on behalf of shangerxin May 19, 2026 07:12 View session

Copilot AI reviewed May 19, 2026

View reviewed changes

shangerxin added 2 commits May 19, 2026 15:18

Fix lint issue

980657c

Fix lint issue

31ae701

Copilot AI review requested due to automatic review settings May 19, 2026 07:40

Copilot started reviewing on behalf of shangerxin May 19, 2026 07:40 View session

Copilot AI reviewed May 19, 2026

View reviewed changes

shangerxin added 2 commits May 19, 2026 15:47

Fix lint issue

98e7817

Ignore lint error

4bd3226

Copilot AI review requested due to automatic review settings May 19, 2026 07:50

Copilot started reviewing on behalf of shangerxin May 19, 2026 07:50 View session

Copilot AI reviewed May 19, 2026

View reviewed changes

shangerxin added 5 commits May 21, 2026 10:27

Fix ut error

7024ccd

Merge branch 'main' into erxin/migrate_dynamo_cases

434c908

Merge branch 'main' into erxin/migrate_dynamo_cases

6f2c9aa

Merge branch 'main' into erxin/migrate_dynamo_cases

ff3889f

Merge branch 'main' into erxin/migrate_dynamo_cases

6ab5175

pponikox mentioned this pull request May 28, 2026

[upstream_ut] test/dynamo/test_activation_checkpointing.py::ActivationCheckpointingViaTagsTestsXPU::test_pattern_matcher failed with IndexError #3393

Open

Merge branch 'main' into erxin/migrate_dynamo_cases

c6f1bc2

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Migrate XPU UT files from Upstream PR#3593

Migrate XPU UT files from Upstream PR#3593
shangerxin wants to merge 18 commits into
intel:mainfrom
shangerxin:erxin/migrate_dynamo_cases

shangerxin commented May 10, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

shangerxin commented May 14, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

shangerxin commented May 18, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants


		self.assertParses()

		@skipIf(HAS_XPU_AND_TRITON, "No backend type associated with device type xpu")

Conversation

shangerxin commented May 10, 2026

Summary

Upstream PR

XPU Issues

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

shangerxin commented May 14, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

shangerxin commented May 18, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants