Require tests to be faster than 30s (#471)

Author: oulgen

.github/workflows/test.yml

@@ -103,8 +103,9 @@ jobs:
         run: |
           source .venv/bin/activate
           uv pip install -r requirements.txt
+          uv pip install pytest-timeout
 
       - name: Run Tests
         run: |
           source .venv/bin/activate
-          HELION_INTERPRET=$([[ "${{ matrix.ref-mode }}" == "ref-eager" ]] && echo "1") pytest
+          HELION_INTERPRET=$([[ "${{ matrix.ref-mode }}" == "ref-eager" ]] && echo "1") pytest --timeout=30

test/test_autotuner.py

@@ -124,6 +124,7 @@ def add(a, b):
         torch.testing.assert_close(add(*args), sum(args))
         torch.testing.assert_close(add(*args), sum(args))
 
+    @skipIfRocm("too slow on rocm")
     def test_random_search(self):
         args = (
             torch.randn([512, 512], device=DEVICE),
@@ -134,6 +135,7 @@ def test_random_search(self):
         fn = bound_kernel.compile_config(best)
         torch.testing.assert_close(fn(*args), args[0] @ args[1], rtol=1e-2, atol=1e-1)
 
+    @skipIfRocm("too slow on rocm")
     def test_differential_evolution_search(self):
         args = (
             torch.randn([512, 512], device=DEVICE),

test/test_indexing.py

@@ -628,6 +628,7 @@ def kernel(buf: torch.Tensor, zeros: torch.Tensor) -> torch.Tensor:
         torch.testing.assert_close(result, expected)
 
     @skipIfRocm("failure on rocm")
+    @unittest.skip("takes 5+ minutes to run")
     def test_1d_indexed_value_from_slice(self):
         """buf2[i] = buf[:] - Assign slice to indexed value"""
 

test/test_views.py

@@ -9,6 +9,7 @@
 from helion._testing import RefEagerTestBase
 from helion._testing import TestCase
 from helion._testing import code_and_output
+from helion._testing import skipIfRocm
 import helion.language as hl
 
 
@@ -33,6 +34,7 @@ def softmax(x: torch.Tensor) -> torch.Tensor:
         )
         self.assertExpectedJournal(code)
 
+    @skipIfRocm("too slow on rocm")
     def test_softmax_view_reshape(self):
         @helion.kernel(config={"block_size": 1})
         def softmax(x: torch.Tensor) -> torch.Tensor: