Merge branch 'NVIDIA:main' into main

Author: rluo8

.github/workflows/test-wheel-linux.yml

@@ -267,7 +267,7 @@ jobs:
         run: |
           pip install pyperf
           pushd cuda_bindings/benchmarks
-          python run_pyperf.py --fast --loops 1 --min-time 1
+          python run_pyperf.py --fast --min-time 1
           popd
 
       - name: Run cuda.core tests

cuda_bindings/benchmarks/pixi.toml

@@ -54,7 +54,7 @@ source = { features = ["cu13", "cu13-source", "bench", "cpp-bench", "dev", "bind
 cmd = ["python", "$PIXI_PROJECT_ROOT/run_pyperf.py"]
 
 [target.linux.tasks.bench-smoke-test]
-cmd = ["python", "$PIXI_PROJECT_ROOT/run_pyperf.py", "--fast", "--loops", "1", "--min-time", "0"
+cmd = ["python", "$PIXI_PROJECT_ROOT/run_pyperf.py", "--fast", "--min-time", "1"
 ]
 
 [target.linux.tasks.bench-legacy]

cuda_bindings/cuda/bindings/_test_helpers/arch_check.py

@@ -8,6 +8,7 @@
 import pytest
 
 from cuda.bindings import nvml
+from cuda.bindings._internal.utils import FunctionNotFoundError as NvmlSymbolNotFoundError
 
 
 @cache
@@ -49,9 +50,10 @@ def unsupported_before(device: int, expected_device_arch: nvml.DeviceArch | str
 
         try:
             yield
-        except nvml.NotSupportedError:
-            # The API call raised NotSupportedError, so we skip the test, but
-            # don't fail it
+        except (nvml.NotSupportedError, nvml.FunctionNotFoundError, NvmlSymbolNotFoundError):
+            # The API call raised NotSupportedError, NVML status FunctionNotFoundError,
+            # or NvmlSymbolNotFoundError (symbol absent from the loaded NVML DLL), so we
+            # skip the test but don't fail it
             pytest.skip(
                 f"Unsupported call for device architecture {nvml.DeviceArch(device_arch).name} "
                 f"on device '{nvml.device_get_name(device)}'"

cuda_core/tests/system/test_system_device.py

@@ -586,7 +586,7 @@ def test_clock():
             with unsupported_before(device, DeviceArch.MAXWELL):
                 try:
                     offsets = clock.get_offsets(pstate)
-                except system.InvalidArgumentError:
+                except (system.InvalidArgumentError, system.NotFoundError):
                     pass
                 else:
                     assert isinstance(offsets, system.ClockOffsets)