[ET-VK] Pass detailed op information to event tracer

[SS-JIA]

Stack from ghstack (oldest at bottom):

• [ET-VK][test-utils] Implement submodule extraction utilities #16267
• -> [ET-VK] Pass detailed op information to event tracer #16266

Context

Currently, when using the event tracer API to log shader execution times, only the shader name is recorded as the event name. However, this provides very minimal context to use when interpreting profiling data. For example, if we see that a convolution shader is running slow, it's is impossible to know from the profiling data alone what the input/output sizes were, what convolution parameters (i.e. stride/padding/dilation) were used, etc.

Changes

This diff makes it so that for each shader dispatch, a JSON is recorded as the event name instead which contains the complete details of all the arguments the operator was invoked with.

The JSON will look something like

    {
      "name": "aten.where.self",
      "args": [
        {
          "type": "TENSOR",
          "value_ref": 25,
          "dtype": "Bool",
          "sizes": [
            1,
            1,
            1,
            8
          ],
          "storage": "TEXTURE_3D",
          "packed_dim": 2
        },
        {
          "type": "TENSOR",
          "value_ref": 30,
          "dtype": "Float",
          "sizes": [
            1,
            6,
            43,
            8
          ],
          "storage": "TEXTURE_3D",
          "packed_dim": 2
        },
        {
          "type": "TENSOR",
          "value_ref": 32,
          "dtype": "Float",
          "sizes": [
            1,
            6,
            43,
            8
          ],
          "storage": "TEXTURE_3D",
          "packed_dim": 2
        },
        {
          "type": "TENSOR",
          "value_ref": 33,
          "dtype": "Float",
          "sizes": [
            1,
            6,
            43,
            8
          ],
          "storage": "TEXTURE_3D",
          "packed_dim": 2
        }
      ]
    }

Then, when processing the profiling data, the JSON can be post-processed to provide useful information to contextualize the shader execution times, for example:

• Memory Throughput
• GFLOPS

Differential Revision: D84646748

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/16266

• 📄 Preview Python docs built from this PR

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

❌ 1 New Failure, 4 Unrelated Failures

As of commit 5286b58 with merge base 5d40a3a ():

NEW FAILURE - The following job has failed:

• Test CUDA Builds / export-model-cuda-artifact (openai, whisper-large-v3-turbo, quantized-int4-weight-only) / linux-job (gh)
  RuntimeError: Command docker exec -t c1a672a5e947f37f20f9b6e739d3f2ec1e06ccb90624f823675e834aa15b1d28 /exec failed with exit code 1

FLAKY - The following jobs failed but were likely due to flakiness present on trunk:

• pull / test-binary-size-linux-gcc / linux-job (gh) (similar failure)
• pull / test-setup-linux-gcc / linux-job (gh) (similar failure)

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-openvino-linux / linux-job (gh) (trunk failure)

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.

[github-actions]

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.