Add nxp backend profiling support

[irtrukhina]

Summary

Add profiling support for the NXP backend.

Test plan

All CI tests passed including new test for the profiling feature.

cc @robert-kalmar @JakeStevens @digantdesai

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/19225

• 📄 Preview Python docs built from this PR

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

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

• [ROCm] MI350 CI jobs will have longer queue times due to CI migration

⏳ 7 Pending, 1 Unrelated Failure

As of commit 6031ad9 with merge base 2b1ccd4 ():

BROKEN TRUNK - The following job failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

• pull / test-llama-runner-qnn-linux (fp32, qnn_16a16w, qnn) / linux-job (gh) (trunk failure)

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

[robert-kalmar]

Review (Part 1/x)

[robert-kalmar]

Please fix the linting errors and look at our comments. For information how to run the lintrunner see https://github.com/pytorch/executorch/blob/main/CONTRIBUTING.md .

In short:

$ lintrunner init
$ lintrunner --merge-base-with <reference to last commit outside of your contribution>

[robert-kalmar]

Internal CI build: https://bamboo3.sw.nxp.com/browse/MLTECE-EXIGH146

[linux-foundation-easycla]

• ❌ - login: @irtrukhina / name: Irina Korchakova. The commit (15dbcbc, 564d1e2, 6031ad9, ab76ed1, b14a37d, ba5b705) is not authorized under a signed CLA. Please click here to be authorized. For further assistance with EasyCLA, please visit our EasyCLA portal and chat with our support bot.

[MartinPavella]

I don't really understand which nodes is this list supposed to contain. What does the comment mean? Are there any single input nodes which shouldn't be in this list?
Also, is this list complete? Or will we ever need to add operators to it?

[irtrukhina]

I created this list from the output of ./neutron-converter --show-kernel-kinds --target imxrt700. I selected all single-input kernels that can be executed in parallel with the same input.
As I understand, no additional kernels are expected for Neutron-C. If that is correct, this should be a complete list for Neutron-C.
The comment means that if two or more kernels from the list are mapped to the same TFLite node, this is incorrect. In such cases, we need to further analyze and filter the results to keep only one Neutron node that corresponds to the TFLite node name.

[irtrukhina]

@robert-kalmar, can you please rerun the pending checks?

[irtrukhina]

Again, a conflict? Sorry, I didn’t notice that. It was perfectly fine yesterday :(