[ET-VK][testing] Create dedicated test binary for pointwise convolutions

[SS-JIA]

Stack from ghstack (oldest at bottom):

• [ET-VK][qconv] Add flexible layout impl for quantized pointwise conv #17221
• -> [ET-VK][testing] Create dedicated test binary for pointwise convolutions #17220
• [ET-VK][qconv] Add layout-agnostic general shader for quantized conv #17219
• [ET-VK][quantization] Add layout-flexible clone for int8x4 tensors #17171
• [ET-VK][qconv] Add layout-flexible impl of quantized depthwise conv2d #17108
• [ET-VK][ez] Implement helper functions to get fastest moving dim #17107
• [ET-VK][quantization] Implement layout-flexible quantize/dequantize operators #17106
• [ET-VK] Add alignment fields to PackedDimInfo for padded size calculation #17170
• [ET-VK][testing] Add per-shader timing breakdown to benchmark output #17105

This commit creates a dedicated test binary for pointwise (1x1) convolutions
(test_q8ta_conv2d_pw), separating them from the general 2D convolution tests.
Here are the key changes:

What Changed

1. New Test Binary: test_q8ta_conv2d_pw.cpp (591 lines)

   • Dedicated test file focusing exclusively on pointwise convolutions (kernel
     size 1x1)
   • Contains 9 test configurations ranging from accuracy tests to performance
     cases:
     • Accuracy tests: Various channel configurations (32→3, 64→32, 96→64, 13→7,
       80→40) with different spatial dimensions
     • Performance tests: Larger configurations (160→480, 22→48, 48→48, 128→128)
       exceeding the 100-dim reference limit
   • Tests all combinations of:
     • Storage types: Texture3D, Buffer
     • Int8 memory layouts: 4C1W, 4W4C, 4C
   • Also tests legacy 4W4C implementation via impl_selector="legacy_4w4c"
   • Includes full reference implementation for numerical correctness
     verification
   • Custom FLOP calculator for performance measurements

2. Removed from test_q8ta_conv2d.cpp (44 lines deleted)

   • Removed 6 pointwise convolution configurations that are now covered by the
     new dedicated binary
   • General conv2d tests now focus solely on kernels > 1x1 (3x3, 5x5, etc.)

3. Build System Updates

   • Added test_q8ta_conv2d_pw target to:
     • targets.bzl (Buck2)
     • CMakeLists.txt (CMake)
   • Both fbcode and xplat paths updated (files are mirrored)

4. CI/Workflow Integration

   • Updated executorch_vulkan_eureka_unit_tests.sky to include the new test
     binary in on-device testing workflow

Why This Separation?

Pointwise convolutions (1x1 kernels) are a distinct optimization target with
different performance characteristics than general convolutions. Separating them
enables:

• Focused performance iteration on pointwise-specific shaders
• Cleaner test organization
• Faster test runs when only testing one convolution type

Differential Revision: D92307251

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/17220

• 📄 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, 5 Unrelated Failures

As of commit 4752b4d with merge base 477867a ():

NEW FAILURES - The following jobs have failed:

• pull / test-vulkan-operators-linux / linux-job (gh)
  RuntimeError: Command docker exec -t 800060064a77aaac857ae828c237e37303b795685db4dff819db21fec3d71b69 /exec failed with exit code 134
• pull / unittest-editable / macos / macos-job (gh)
  backends/xnnpack/test/recipes/test_xnnpack_recipes.py::TestXnnpackRecipes::test_int8_static_quant_recipe
• Test CUDA Builds / export-model-cuda-artifact (openai, whisper-small, quantized-int4-tile-packed) / linux-job (gh)
  RuntimeError: Command docker exec -t 3fbb82184ad2451f09273f3cc47effdc714137f4ffa1872345f614a5e1d66b11 /exec failed with exit code 1
• Test CUDA Builds / export-model-cuda-artifact (openai, whisper-small, quantized-int4-weight-only) / linux-job (gh)
  RuntimeError: Command docker exec -t 34cae9d4d71a0d0b18e93a8cb63d5dbc4847553f40d0ed23f7b788bea2983919 /exec failed with exit code 1

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

• pull / test-models-linux (ic4, portable, linux.4xlarge.memory) / linux-job (gh) (detected as infra flaky with no log or failing log classifier)
• pull / test-models-linux (llama3_2_vision_encoder, portable, linux.4xlarge.memory) / linux-job (gh) (detected as infra flaky with no log or failing log classifier)
• pull / test-models-linux (w2l, portable, linux.4xlarge.memory) / linux-job (gh) (detected as infra flaky with no log or failing log classifier)

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

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

• pull / test-samsung-models-linux / linux-job (gh) (trunk failure)
  ##[error]The operation was canceled.
• pull / test-samsung-quantmodels-linux / linux-job (gh) (trunk failure)
  ##[error]The operation was canceled.

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.