NXP backend: Add edge dialect pass to remove useless as_strided_copy nodes.

[MartinPavella]

Summary

This PR integrates the RemoveUselessOpPass edge dialect pass from the Samsung backend into the NXP backend. This feature is required to support an internal model.

Test plan

Tested internally by NXP on a private model.

cc @robert-kalmar

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/16489

• 📄 Preview Python docs built from this PR

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

❌ 4 New Failures, 1 Unrelated Failure

As of commit 276e961 with merge base e847384 ():

NEW FAILURES - The following jobs have failed:

• pull / unittest-arm-backend-with-no-deps (test_pytest_models_tosa) / linux-job (gh)
  RuntimeError: Command docker exec -t de8d3d0e72704a81770c01761908515389c3a6ec6211f52ef3e987c95eee58de /exec failed with exit code 127
• pull / unittest-arm-backend-with-no-deps (test_pytest_ops_no_target) / linux-job (gh)
  RuntimeError: Command docker exec -t 9f161d561b9952c076a347a01d202ba494177a25c5fe9582b68bc21d77387d93 /exec failed with exit code 127
• pull / unittest-arm-backend-with-no-deps (test_pytest_ops_tosa) / linux-job (gh)
  RuntimeError: Command docker exec -t 94600f966c4a97506a997f4c7e2bcb16e35c5fef2a798981ffd9b4a6f163d770 /exec failed with exit code 127
• pull / unittest-arm-backend-with-no-deps (test_run_tosa) / linux-job (gh)
  RuntimeError: Command docker exec -t 042ca0839ace61290370961b483601a2df8a462aa3469c35013c3384bf037766 /exec failed with exit code 127

UNSTABLE - The following job is marked as unstable, possibly due to flakiness on trunk:

• pull / android / run-emulator (gh) (#16137)
  Timeout waiting for emulator to boot.

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

[robert-kalmar]

@MartinPavella , you mentioned the RemoveUselessOpPass handled specifically the case of mean followed by the as_strided_copy, but seems it handles more "useless" op cases - https://github.com/pytorch/executorch/blob/main/backends/samsung/_passes/remove_useless_ops.py#L15

[MartinPavella]

Now that you mention it, I thought about it a bit more. For our use case we actually mustn't remove the other "useless ops". Since we need to run this pass before partitioning, some of the ops (like dim_order_ops._to_dim_order_copy) are not useless at all, and are actually vital in case they don't get delegated.

Therefore, I will create our own pass based on this one, to only remove the as_strided_copy for now.

[MartinPavella]

Done

[robert-kalmar]

@Jiseong-oh, we also need a pass to remove the as_strided_copy preceeded by the mean, what handled by the executorch.backends.samsung._passes.remove_useless_ops.RemoveUselessOpPass. Do you mind to reuse it in NXP backend?

[robert-kalmar]

@Jiseong-oh, we also need a pass to remove the as_strided_copy preceeded by the mean, what handled by the executorch.backends.samsung._passes.remove_useless_ops.RemoveUselessOpPass. Do you mind to reuse it in NXP backend?

@Jiseong-oh , based on the feedback from Martin, the question is not anymore relevant.

[robert-kalmar]

Code looks fine to me.
Let the CI's finish.