[CI] Add aarch64 operator benchmark (#165585)

Running on Graviton4
Skip ConvTranspose1d benchmarks if PyTorch is compiled with ACL, due to https://github.com/pytorch/pytorch/issues/165654
Pull Request resolved: https://github.com/pytorch/pytorch/pull/165585
Approved by: https://github.com/huydhn

.github/workflows/operator_benchmark.yml

@@ -52,3 +52,27 @@ jobs:
       docker-image: ${{ needs.x86-opbenchmark-build.outputs.docker-image }}
       test-matrix: ${{ needs.x86-opbenchmark-build.outputs.test-matrix }}
     secrets: inherit
+
+  aarch64-opbenchmark-build:
+    if: github.repository_owner == 'pytorch'
+    name: aarch64-opbenchmark-build
+    uses: ./.github/workflows/_linux-build.yml
+    with:
+      build-environment: linux-jammy-aarch64-py3.10
+      runner: linux.arm64.m7g.4xlarge
+      docker-image-name: ci-image:pytorch-linux-jammy-aarch64-py3.10-gcc11
+      test-matrix: |
+        { include: [
+          { config: "cpu_operator_benchmark_short", shard: 1, num_shards: 1, runner: "linux.arm64.m8g.4xlarge" },
+        ]}
+    secrets: inherit
+
+  aarch64-opbenchmark-test:
+    name: aarch64-opbenchmark-test
+    uses: ./.github/workflows/_linux-test.yml
+    needs: aarch64-opbenchmark-build
+    with:
+      build-environment: linux-jammy-aarch64-py3.10
+      docker-image: ${{ needs.aarch64-opbenchmark-build.outputs.docker-image }}
+      test-matrix: ${{ needs.aarch64-opbenchmark-build.outputs.test-matrix }}
+    secrets: inherit

benchmarks/operator_benchmark/pt/conv_test.py

@@ -38,12 +38,16 @@ class ConvTranspose1dBenchmark(op_bench.TorchBenchmarkBase):
 op_bench.generate_pt_test(
     configs.conv_1d_configs_short + configs.conv_1d_configs_long, Conv1dBenchmark
 )
-op_bench.generate_pt_test(
-    configs.convtranspose_1d_configs_short
-    + configs.conv_1d_configs_short
-    + configs.conv_1d_configs_long,
-    ConvTranspose1dBenchmark,
-)
+
+
+if not torch.backends.mkldnn.is_acl_available():
+    # convtranpose1d crashes with ACL, see https://github.com/pytorch/pytorch/issues/165654
+    op_bench.generate_pt_test(
+        configs.convtranspose_1d_configs_short
+        + configs.conv_1d_configs_short
+        + configs.conv_1d_configs_long,
+        ConvTranspose1dBenchmark,
+    )
 
 
 """