enforce the right CC is passed to NVRTC

NVIDIA · Dec 8, 2024 · b1f07a3 · b1f07a3
1 parent f3cc6bd
commit b1f07a3
Show file tree

Hide file tree

Showing 2 changed files with 5 additions and 3 deletions.
diff --git a/cuda_core/tests/conftest.py b/cuda_core/tests/conftest.py
@@ -18,7 +18,7 @@
 import pytest
 
 from cuda.core.experimental import Device, _device
-from cuda.core.experimental._utils import CUDAError, handle_return
+from cuda.core.experimental._utils import handle_return
 
 
 @pytest.fixture(scope="session", autouse=True)

diff --git a/cuda_core/tests/test_program.py b/cuda_core/tests/test_program.py
@@ -9,7 +9,7 @@
 import pytest
 
 from cuda import cuda, nvrtc
-from cuda.core.experimental import Program
+from cuda.core.experimental import Device, Program
 from cuda.core.experimental._module import Kernel, ObjectCode
 
 
@@ -46,7 +46,9 @@ def test_program_init_invalid_code_format():
 def test_program_compile_valid_target_type():
     code = 'extern "C" __global__ void my_kernel() {}'
     program = Program(code, "c++")
-    object_code = program.compile("ptx")
+    arch = "".join(str(i) for i in Device().compute_capability)
+    object_code = program.compile("ptx", options=(f"-arch=compute_{arch}",))
+    print(object_code._module.decode())
     kernel = object_code.get_kernel("my_kernel")
     assert isinstance(object_code, ObjectCode)
     assert isinstance(kernel, Kernel)