def test_nvml_nvlink_counters(ngpus, handles, counter, control, driver):
if driver > 450.0:
pytest.xfail(XFAIL_LEGACY_NVLINK_MSG)
reset = 0
for i in range(ngpus):
for j in range(pynvml.NVML_NVLINK_MAX_LINKS):
assert (
pynvml.nvmlDeviceResetNvLinkUtilizationCounter(handles[i], j, counter)
== pynvml.NVML_SUCCESS
)
pynvml.nvmlDeviceSetNvLinkUtilizationControl(
handles[i], j, counter, control, reset
)
countdict = pynvml.nvmlDeviceGetNvLinkUtilizationCounter(
handles[i], j, counter
)
ctl = pynvml.nvmlDeviceGetNvLinkUtilizationControl(handles[i], j, counter)
assert countdict["rx"] >= 0
assert countdict["tx"] >= 0
assert ctl == control
assert (
pynvml.nvmlDeviceFreezeNvLinkUtilizationCounter(
handles[i], j, counter, 1
)
== pynvml.NVML_SUCCESS
)
assert (
pynvml.nvmlDeviceFreezeNvLinkUtilizationCounter(
handles[i], j, counter, 0
)
== pynvml.NVML_SUCCESS
)
def test_nvml_nvlink_counters(ngpus, handles, counter, control):
reset = 0
for i in range(ngpus):
for j in range(pynvml.NVML_NVLINK_MAX_LINKS):
assert pynvml.nvmlDeviceResetNvLinkUtilizationCounter(
handles[i], j, counter) == pynvml.NVML_SUCCESS
pynvml.nvmlDeviceSetNvLinkUtilizationControl(
handles[i], j, counter, control, reset)
countdict = pynvml.nvmlDeviceGetNvLinkUtilizationCounter(
handles[i], j, counter)
ctl = pynvml.nvmlDeviceGetNvLinkUtilizationControl(
handles[i], j, counter)
assert countdict['rx'] >= 0
assert countdict['tx'] >= 0
assert ctl == control
assert pynvml.nvmlDeviceFreezeNvLinkUtilizationCounter(
handles[i], j, counter, 1) == pynvml.NVML_SUCCESS
assert pynvml.nvmlDeviceFreezeNvLinkUtilizationCounter(
handles[i], j, counter, 0) == pynvml.NVML_SUCCESS