[ET-VK][custom_ops] Standardize test case label format across binaries

[SS-JIA]

Stack from ghstack (oldest at bottom):

• [ET-VK][custom_ops] Standardize binary filenames to test_*.cpp #19571
• -> [ET-VK][custom_ops] Standardize test case label format across binaries #19570
• [ET-VK][custom_ops] Probe-then-scale adaptive dispatch stacking in test framework #19569

Previously the test case labels printed by each custom_ops test binary used wildly inconsistent formats. Examples from a recent on-device pass:

ACCU  conv1d_dw [1,16,64] K=3 s=1 p=1 d=1  Tex(HP) f32
ACCU  I=144  Buf(4W)
correctness_1_64_32_Buffer_Float
small_1d_linear_weight
ACCU  In=1,16,4,4  r=2  same_qp  [4W4C->4W4C]
ACCU  32->3  I=64,64  g=1  k=3  Buf(4C1W) [general]

Standardize all 21 binaries on:

[ACCU/PERF]  [INPUT->OUTPUT dtype]  [shape data + inline op params]  [Storage(layout)]  [optional suffix]

Concrete examples after this commit:

ACCU  f32->f32  [1,16,64] k3 s1 p1 d1                Tex(HP)
ACCU  i8->i8    [144]+[144]                          Buf(4W)
ACCU  f32->f32  [1,64]x[32,64]                       Buf(WP)
ACCU  i32->f32  [1,8] lw                             Tex(WP)
ACCU  i8->i8    [1,16,4,4] r=2                       Buf(4W4C)->Buf(4W4C)  [same_qp]
ACCU  f32->f32  [1,3,64,64]x[32,3,3,3] s1 p1 d1 g1   Buf(4C1W)  [general]

Adds three helpers to test/custom_ops/utils.h:

• dtype_short(vkapi::ScalarType) -> "f32" / "f16" / "i8" / "i32"
• shape_bracket(std::vector<int64_t>) -> "[N,C,H,W]" form
• make_test_label(prefix, in_dtype, out_dtype, shape_str, storage_str, suffix) assembles the four sections with two-space separators and an optional bracketed suffix.

Each binary's TestCase-builder helper now invokes make_test_label rather than rolling its own ostringstream. Op-specific shape detail (kernel/stride/padding/dilation/groups for conv, ratio for pixel_shuffle, output_padding for transposed conv, layout-transition arrows for clone/qdq) is constructed inline since it varies per op. Information that was load-bearing in the old labels (impl_selector tags, +bias/no_bias flags, same_qp/diff_qp markers, const_b for binary ops, multi-stage Buf->Buf transitions) is preserved as a bracketed suffix.

Two minor format deviations worth flagging:

• choose_qparams_per_row produces two outputs (scale tensor + zero_point tensor). Encoded as out_dtype = "f32,i8" -- a comma-separated tuple after the arrow -- rather than forcing a single token.
• test_embedding_q4gsw loses the free-form prose labels (small_1d_linear_weight, llama_3_2_1b_prefill_*). Replaced with structural shape info + an lw / scales-dtype suffix; readers can still derive the same context but it is less greppable as a "story".

Differential Revision: D105059941

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/19570

• 📄 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:

• Run pull request jobs on OSDC runners in shadow mode

❌ 4 Pending, 2 Unrelated Failures, 1 Unclassified Failure

As of commit 0d3c455 with merge base 1992bdd ():

UNCLASSIFIED FAILURE - DrCI could not classify the following job because the workflow did not run on the merge base. The failure may be pre-existing on trunk or introduced by this PR:

• Check Labels (gh) (this job did not run on the merge base, so DrCI cannot tell whether the failure is pre-existing)

FLAKY - The following job failed but was likely due to flakiness present on trunk:

• pull / unittest-editable / windows / windows-job (gh) (similar failure)
  backends/xnnpack/test/recipes/test_xnnpack_recipes.py::TestXnnpackRecipes::test_8a4w_recipe

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

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

• pull / unittest / windows / windows-job (gh) (trunk failure)
  backends/xnnpack/test/recipes/test_xnnpack_recipes.py::TestXnnpackRecipes::test_8a4w_recipe

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