def main():
#Make sure logging stuff is bootstrapped
try:
option_parser.parse_options()
option_parser.options.no_logging = True #Don't log anything
heHandle = None
heAppRunner = None
dcgm_structs._LoadDcgmLibrary()
if g_embeddedMode:
host = 0
else:
#Start host engine
heAppRunner = apps.NvHostEngineApp()
heAppRunner.start(timeout=1000000000)
time.sleep(2.0)
host = "127.0.0.1"
heHandle = dcgm_agent.dcgmInit()
pssObj = ProcessStatsStress(g_embeddedMode, heHandle)
pssObj.Run()
del (pssObj) #Force destructor
heAppRunner.wait()
except Exception as e:
raise
finally:
apps.AppRunner.clean_all()
if heHandle is not None:
dcgm_agent.dcgmShutdown()
assert config_values[x].mComputeMode == expected_compute_mode, "The compute mode value for gpuID %d is incorrect."\
" Returned: %d Expected: %d" \
% (x, config_values[x].mComputeMode, expected_compute_mode)
assert config_values[x].mEccMode == expected_ecc, "The ecc mode value for gpuID %d is incorrect."\
" Returned: %d Expected: %d" \
% (x, config_values[x].mEccMode, expected_ecc)
pass
ret = dcgm_agent.dcgmStatusDestroy(status_handle)
assert (ret == dcgm_structs.DCGM_ST_OK
), "Failed to remove status handler, error: %s" % ret
dcgm_structs._LoadDcgmLibrary()
handle = dcgm_agent.dcgmInit()
devices = dcgm_agent.dcgmGetAllDevices(handle)
validDevices = list()
for x in devices:
fvSupported = dcgm_agent_internal.dcgmGetLatestValuesForFields(
handle, x, [
dcgm_fields.DCGM_FI_DEV_RETIRED_DBE,
])
if (fvSupported[0].value.i64 != dcgmvalue.DCGM_INT64_NOT_SUPPORTED):
validDevices.append(x)
if (len(validDevices) == 0):
print "Can only run if at least one GPU with ECC is present"
sys.exit(1)