def _InstallAndAuthenticateVm(vm):
"""Install SHOC, ensure correct GPU state, and authenticate the VM for ssh.
Args:
vm: vm to operate on.
"""
vm.Install('shoc_benchmark_suite')
nvidia_driver.SetAndConfirmGpuClocks(vm)
vm.AuthenticateVm() # Configure ssh between vms for MPI
def _PrepareHpcg(vm):
"""Install HPCG on a single vm.
Args:
vm: vm to operate on
"""
logging.info('Installing HPCG on %s', vm)
vm.Install('hpcg')
vm.AuthenticateVm()
nvidia_driver.SetAndConfirmGpuClocks(vm)
def Run(benchmark_spec):
"""Sets the GPU clock speed and runs the CUDA PCIe benchmark.
Args:
benchmark_spec: The benchmark specification. Contains all data that is
required to run the benchmark.
Returns:
A list of sample.Sample objects.
"""
vm = benchmark_spec.vms[0]
# Note: The clock speed is set in this function rather than Prepare()
# so that the user can perform multiple runs with a specified
# clock speed without having to re-prepare the VM.
nvidia_driver.SetAndConfirmGpuClocks(vm)
num_iterations = FLAGS.gpu_pcie_bandwidth_iterations
mode = FLAGS.gpu_pcie_bandwidth_mode
transfer_size_range = FLAGS.gpu_pcie_bandwidth_transfer_sizes
raw_results = []
metadata = {}
metadata.update(cuda_toolkit.GetMetadata(vm))
metadata['num_iterations'] = num_iterations
metadata['mode'] = mode
if mode == 'range':
metadata['range_start'] = transfer_size_range[0]
metadata['range_stop'] = transfer_size_range[1]
metadata['range_step'] = transfer_size_range[2]
run_command = ('%s/extras/demo_suite/bandwidthTest --device=all' %
metadata['cuda_toolkit_home'])
if mode == 'range':
run_command += (
' --mode=range --start={0} --end={1} --increment={2}'.format(
transfer_size_range[0], transfer_size_range[1],
transfer_size_range[2]))
for i in range(num_iterations):
stdout, _ = vm.RemoteCommand(run_command, should_log=True)
raw_results.append(_ParseOutputFromSingleIteration(stdout))
if 'device_info' not in metadata:
metadata['device_info'] = _ParseDeviceInfo(stdout)
return _CalculateMetricsOverAllIterations(raw_results, metadata)