Skip to content

Commit 69009b0

Browse files
rwgkrwgk
committed
Isolate independent inner test cases
1 parent 1344064 commit 69009b0

3 files changed

Lines changed: 105 additions & 52 deletions

File tree

‎cuda_bindings/tests/nvml/test_compute_mode.py‎

Lines changed: 10 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -20,15 +20,23 @@
2020
@pytest.mark.skipif(sys.platform == "win32", reason="Test not supported on Windows")
2121
def test_compute_mode_supported_nonroot(all_devices, subtests):
2222
for device in all_devices:
23-
with subtests.test(device=device):
23+
device_index = nvml.device_get_index(device)
24+
original_compute_mode = None
25+
with subtests.test(device_index=device_index, compute_mode_api="get_compute_mode"):
2426
with unsupported_before(device, None):
2527
original_compute_mode = nvml.device_get_compute_mode(device)
28+
if original_compute_mode is None:
29+
continue
2630

27-
for cm in COMPUTE_MODES:
31+
for cm in COMPUTE_MODES:
32+
with subtests.test(device_index=device_index, compute_mode=cm.name):
2833
try:
2934
nvml.device_set_compute_mode(device, cm)
3035
except nvml.NoPermissionError:
3136
pytest.skip("Insufficient permissions to set compute mode")
37+
except nvml.NvmlError:
38+
nvml.device_set_compute_mode(device, original_compute_mode)
39+
raise
3240
nvml.device_set_compute_mode(device, original_compute_mode)
3341
assert original_compute_mode == nvml.device_get_compute_mode(device), (
3442
"Compute mode shouldn't have changed"

‎cuda_bindings/tests/nvml/test_pynvml.py‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -285,8 +285,8 @@ def test_device_get_pcie_throughput(ngpus, handles, subtests):
285285
) # Link is supported on this device
286286
def test_device_get_nvlink_capability(ngpus, handles, cap_type, subtests):
287287
for i in range(ngpus):
288-
with subtests.test(i=i):
289-
for j in range(nvml.NVLINK_MAX_LINKS):
288+
for j in range(nvml.NVLINK_MAX_LINKS):
289+
with subtests.test(device_index=i, link_index=j):
290290
# By the documentation, this should be supported on PASCAL or newer,
291291
# but this also seems to fail on newer.
292292
with unsupported_before(handles[i], None):

‎cuda_core/tests/system/test_system_device.py‎

Lines changed: 93 additions & 48 deletions
Original file line numberDiff line numberDiff line change
@@ -98,12 +98,21 @@ def test_device_cpu_affinity(subtests):
9898
@pytest.mark.skipif(helpers.IS_WSL or helpers.IS_WINDOWS, reason="Device attributes not supported on WSL or Windows")
9999
def test_affinity(subtests):
100100
for device in system.Device.get_all_devices():
101-
with subtests.test(device=device):
102-
for scope in typing.AffinityScope.__members__.values():
101+
for scope in typing.AffinityScope.__members__.values():
102+
with subtests.test(
103+
device_index=device.index,
104+
affinity_scope=scope.value,
105+
affinity_api="get_cpu_affinity",
106+
):
103107
with unsupported_before(device, typing.DeviceArch.KEPLER):
104108
affinity = device.get_cpu_affinity(scope)
105109
assert isinstance(affinity, list)
106110

111+
with subtests.test(
112+
device_index=device.index,
113+
affinity_scope=scope.value,
114+
affinity_api="get_memory_affinity",
115+
):
107116
affinity = device.get_memory_affinity(scope)
108117
assert isinstance(affinity, list)
109118

@@ -410,8 +419,11 @@ def test_field_values(subtests):
410419
@pytest.mark.skipif(helpers.IS_WSL or helpers.IS_WINDOWS, reason="Device attributes not supported on WSL or Windows")
411420
def test_get_all_devices_with_cpu_affinity(subtests):
412421
for i in range(multiprocessing.cpu_count()):
413-
with subtests.test(i=i):
414-
for device in system.Device.get_all_devices_with_cpu_affinity(i):
422+
devices = []
423+
with subtests.test(cpu_index=i, affinity_api="get_all_devices_with_cpu_affinity"):
424+
devices = list(system.Device.get_all_devices_with_cpu_affinity(i))
425+
for device in devices:
426+
with subtests.test(cpu_index=i, device_index=device.index):
415427
with unsupported_before(device, DeviceArch.KEPLER):
416428
affinity = device.get_cpu_affinity()
417429
assert isinstance(affinity, list)
@@ -566,8 +578,8 @@ def test_auto_boosted_clocks_enabled(subtests):
566578

567579
def test_clock(subtests):
568580
for device in system.Device.get_all_devices():
569-
with subtests.test(device=device):
570-
for clock_type in typing.ClockType:
581+
for clock_type in typing.ClockType:
582+
with subtests.test(device_index=device.index, clock_type=clock_type.value):
571583
clock = device.get_clock(clock_type)
572584
assert isinstance(clock, _device.ClockInfo)
573585

@@ -628,44 +640,45 @@ def test_clock_event_reasons(subtests):
628640

629641
def test_fan(subtests):
630642
for device in system.Device.get_all_devices():
631-
with subtests.test(device=device):
632-
# The fan APIs are only supported on discrete devices with fans,
633-
# but when they are not available `device.num_fans` returns 0.
634-
if device.num_fans == 0:
643+
# The fan APIs are only supported on discrete devices with fans,
644+
# but when they are not available `device.num_fans` returns 0.
645+
if device.num_fans == 0:
646+
with subtests.test(device_index=device.index):
635647
pytest.skip("Device has no fans to test")
648+
continue
636649

637-
for fan_idx in range(device.num_fans):
638-
with subtests.test(fan_idx=fan_idx):
639-
fan_info = device.get_fan(fan_idx)
640-
assert isinstance(fan_info, _device.FanInfo)
650+
for fan_idx in range(device.num_fans):
651+
with subtests.test(device_index=device.index, fan_index=fan_idx):
652+
fan_info = device.get_fan(fan_idx)
653+
assert isinstance(fan_info, _device.FanInfo)
641654

642-
speed = fan_info.speed
643-
assert isinstance(speed, int)
644-
assert 0 <= speed <= 200
645-
try:
646-
fan_info.speed = 50
647-
except nvml.NoPermissionError as e:
648-
pytest.xfail(f"nvml.NoPermissionError: {e}")
649-
try:
650-
fan_info.speed = speed
655+
speed = fan_info.speed
656+
assert isinstance(speed, int)
657+
assert 0 <= speed <= 200
658+
try:
659+
fan_info.speed = 50
660+
except nvml.NoPermissionError as e:
661+
pytest.xfail(f"nvml.NoPermissionError: {e}")
662+
try:
663+
fan_info.speed = speed
651664

652-
speed_rpm = fan_info.speed_rpm
653-
assert isinstance(speed_rpm, int)
654-
assert speed_rpm >= 0
665+
speed_rpm = fan_info.speed_rpm
666+
assert isinstance(speed_rpm, int)
667+
assert speed_rpm >= 0
655668

656-
target_speed = fan_info.target_speed
657-
assert isinstance(target_speed, int)
658-
assert speed <= target_speed * 2
669+
target_speed = fan_info.target_speed
670+
assert isinstance(target_speed, int)
671+
assert speed <= target_speed * 2
659672

660-
min_, max_ = fan_info.min_max_speed
661-
assert isinstance(min_, int)
662-
assert isinstance(max_, int)
663-
assert min_ <= max_
673+
min_, max_ = fan_info.min_max_speed
674+
assert isinstance(min_, int)
675+
assert isinstance(max_, int)
676+
assert min_ <= max_
664677

665-
control_policy = fan_info.control_policy
666-
assert isinstance(control_policy, typing.FanControlPolicy)
667-
finally:
668-
fan_info.set_default_speed()
678+
control_policy = fan_info.control_policy
679+
assert isinstance(control_policy, typing.FanControlPolicy)
680+
finally:
681+
fan_info.set_default_speed()
669682

670683

671684
def test_cooler(subtests):
@@ -690,32 +703,52 @@ def test_cooler(subtests):
690703

691704
def test_temperature(subtests):
692705
for device in system.Device.get_all_devices():
693-
with subtests.test(device=device):
706+
device_index = device.index
707+
temperature = None
708+
with subtests.test(device_index=device_index, temperature_api="temperature"):
694709
temperature = device.temperature
695710
assert isinstance(temperature, _device.Temperature)
711+
if temperature is None:
712+
continue
696713

714+
with subtests.test(device_index=device_index, temperature_api="get_sensor"):
697715
sensor = temperature.get_sensor()
698716
assert isinstance(sensor, int)
699717
assert sensor >= 0
700718

701-
# By docs, should be supported on KEPLER or newer, but experimentally,
702-
# is also unsupported on other hardware.
703-
with unsupported_before(device, None):
704-
for threshold in list(typing.TemperatureThresholds):
719+
# By docs, should be supported on KEPLER or newer, but experimentally,
720+
# is also unsupported on other hardware.
721+
for threshold in typing.TemperatureThresholds:
722+
with subtests.test(
723+
device_index=device_index,
724+
temperature_api="get_threshold",
725+
threshold=threshold.value,
726+
):
727+
with unsupported_before(device, None):
705728
t = temperature.get_threshold(threshold)
706-
assert isinstance(t, int)
707-
assert t >= 0
729+
assert isinstance(t, int)
730+
assert t >= 0
708731

732+
with subtests.test(device_index=device_index, temperature_api="margin"):
709733
with unsupported_before(device, None):
710734
margin = temperature.margin
711735
assert isinstance(margin, int)
712736
assert margin >= 0
713737

738+
thermals = None
739+
with subtests.test(device_index=device_index, temperature_api="get_thermal_settings"):
714740
with unsupported_before(device, None):
715741
thermals = temperature.get_thermal_settings(typing.ThermalTarget.ALL)
716742
assert isinstance(thermals, _device.ThermalSettings)
743+
if thermals is None:
744+
continue
717745

718-
for i, sensor in enumerate(thermals):
746+
for i, sensor in enumerate(thermals):
747+
with subtests.test(
748+
device_index=device_index,
749+
temperature_api="thermal_sensor",
750+
sensor_index=i,
751+
):
719752
assert isinstance(sensor, _device.ThermalSensor)
720753
assert isinstance(sensor.target, typing.ThermalTarget)
721754
assert isinstance(sensor.controller, typing.ThermalController)
@@ -804,8 +837,14 @@ def test_compute_running_processes(subtests):
804837

805838
def test_nvlink(subtests):
806839
for device in system.Device.get_all_devices():
807-
with subtests.test(device=device), unsupported_before(device, None):
808-
for link in range(device.get_nvlink_count()):
840+
device_index = device.index
841+
link_count = 0
842+
with subtests.test(device_index=device_index, nvlink_api="get_nvlink_count"):
843+
with unsupported_before(device, None):
844+
link_count = device.get_nvlink_count()
845+
846+
for link in range(link_count):
847+
with subtests.test(device_index=device_index, nvlink_api="get_nvlink", link_index=link):
809848
with unsupported_before(device, None):
810849
nvlink_info = device.get_nvlink(link)
811850
assert isinstance(nvlink_info, _device.NvlinkInfo)
@@ -823,7 +862,13 @@ def test_nvlink(subtests):
823862
assert len(version) == 2
824863
assert all(isinstance(i, int) for i in version)
825864

826-
for nvlink_info in device.get_nvlinks():
865+
nvlink_infos = []
866+
with subtests.test(device_index=device_index, nvlink_api="get_nvlinks"):
867+
with unsupported_before(device, None):
868+
nvlink_infos = list(device.get_nvlinks())
869+
870+
for link, nvlink_info in enumerate(nvlink_infos):
871+
with subtests.test(device_index=device_index, nvlink_api="get_nvlinks", link_index=link):
827872
assert isinstance(nvlink_info, _device.NvlinkInfo)
828873

829874
with unsupported_before(device, None):

0 commit comments

Comments
 (0)