Instruments

acmecape

Instrumetnation for the BayLibre ACME cape for power/energy measurment.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
iio_capture : str
Path to the iio-capture binary will be taken from the environment, if not specfied.
host : str

Host name (or IP address) of the ACME cape board.

default: 'baylibre-acme.local'

iio_device : str

default: 'iio:device0'
buffer_size : integer

Size of the capture buffer (in KB).

default: 256

cci_pmu_logger

This instrument allows collecting CCI counter data.

It relies on the pmu_logger.ko kernel driver, the source for which is included with Workload Automation (see inside wlauto/external directory). You will need to build this against your specific kernel. Once compiled, it needs to be placed in the dependencies directory (usually ~/.workload_uatomation/dependencies).

Note

When compling pmu_logger.ko for a new hardware platform, you may need to modify CCI_BASE inside pmu_logger.c to contain the base address of where CCI is mapped in memory on your device.

This instrument relies on trace-cmd instrument to also be enabled. You should enable at least 'bprint' trace event.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
events : list

A list of strings, each representing an event to be counted. The length of the list cannot exceed the number of PMU counters available (4 in CCI-400). If this is not specified, shareable read transactions and snoop hits on both clusters will be counted by default. E.g. ['0x63', '0x83'].

default: ['0x63', '0x6A', '0x83', '0x8A']

event_labels : list
A list of labels to be used when reporting PMU counts. If specified, this must be of the same length as cci_pmu_events. If not specified, events will be labeled “event_<event_number>”.
period : integer

The period (in jiffies) between counter reads.

default: 10

install_module : boolean

Specifies whether pmu_logger has been compiled as a .ko module that needs to be installed by the instrument. (.ko binary must be in /home/docs/.workload_automation/dependencies). If this is set to False, it will be assumed that pmu_logger has been compiled into the kernel, or that it has been installed prior to the invocation of WA.

default: True

coreutil

Measures CPU core activity during workload execution in terms of the percentage of time a number of cores were utilized above the specfied threshold.

This workload generates coreutil.csv report in the workload’s output directory. The report is formatted as follows:

<threshold,1core,2core,3core,4core
18.098132,38.650248000000005,10.736180000000001,3.6809760000000002,28.834312000000001

Interpretation of the result:

  • 38.65% of total time only single core is running above or equal to threshold value
  • 10.736% of total time two cores are running simultaneously above or equal to threshold value
  • 3.6809% of total time three cores are running simultaneously above or equal to threshold value
  • 28.8314% of total time four cores are running simultaneously above or equal to threshold value
  • 18.098% of time all core are running below threshold value.

..note : This instrument doesn’t work on ARM big.LITTLE IKS implementation

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
threshold : integer

Cores with percentage utilization above this value will be considered as “utilized”. This value may need to be adjusted based on the background activity and the intensity of the workload being instrumented (e.g. it may need to be lowered for low-intensity workloads such as video playback).

constraint: 0 < value <= 100

default: 50

cpufreq

Collects dynamic frequency (DVFS) settings before and after workload execution.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
paths : list_of_strs
A list of paths to be pulled from the device. These could be directories as well as files.
use_tmpfs : boolean
Specifies whether tmpfs should be used to cache sysfile trees and then pull them down as a tarball. This is significantly faster then just copying the directory trees from the device directly, bur requres root and may not work on all devices. Defaults to True if the device is rooted and False if it is not.
tmpfs_mount_point : str
Mount point for tmpfs partition used to store snapshots of paths.
tmpfs_size : str

Size of the tempfs partition.

default: '32m'

daq

DAQ instrument obtains the power consumption of the target device’s core measured by National Instruments Data Acquisition(DAQ) device.

WA communicates with a DAQ device server running on a Windows machine (Please refer to DAQ Server Guide) over a network. You must specify the IP address and port the server is listening on in the config file as follows

daq_server_host = '10.1.197.176'
daq_server_port = 45677

These values will be output by the server when you run it on Windows.

You must also specify the values of resistors (in Ohms) across which the voltages are measured (Please refer to DAQ Server Guide). The values should be specified as a list with an entry for each resistor, e.g.:

daq_resistor_values = [0.005, 0.005]

In addition to this mandatory configuration, you can also optionally specify the following:

:daq_labels: Labels to be used for ports. Defaults to ``'PORT_<pnum>'``, where
             'pnum' is the number of the port.
:daq_device_id: The ID under which the DAQ is registered with the driver.
                Defaults to ``'Dev1'``.
:daq_v_range: Specifies the voltage range for the SOC voltage channel on the DAQ
              (please refer to :ref:`daq_setup` for details). Defaults to ``2.5``.
:daq_dv_range: Specifies the voltage range for the resistor voltage channel on
               the DAQ (please refer to :ref:`daq_setup` for details).
               Defaults to ``0.2``.
:daq_sampling_rate: DAQ sampling rate. DAQ will take this many samples each
                    second. Please note that this maybe limitted by your DAQ model
                    and then number of ports you're measuring (again, see
                    :ref:`daq_setup`). Defaults to ``10000``.
:daq_channel_map: Represents mapping from  logical AI channel number to physical
                  connector on the DAQ (varies between DAQ models). The default
                  assumes DAQ 6363 and similar with AI channels on connectors
                  0-7 and 16-23.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
server_host : str

The host address of the machine that runs the daq Server which the instrument communicates with.

default: 'localhost'

server_port : integer

The port number for daq Server in which daq instrument communicates with.

default: 45677

device_id : str

The ID under which the DAQ is registered with the driver.

default: 'Dev1'

v_range : float

Specifies the voltage range for the SOC voltage channel on the DAQ (please refer to DAQ Server Guide for details).

default: 2.5

dv_range : float

Specifies the voltage range for the resistor voltage channel on the DAQ (please refer to DAQ Server Guide for details).

default: 0.2

sampling_rate : integer

DAQ sampling rate. DAQ will take this many samples each second. Please note that this maybe limitted by your DAQ model and then number of ports you’re measuring (again, see DAQ Server Guide)

default: 10000

resistor_values : list (mandatory)
The values of resistors (in Ohms) across which the voltages are measured on each port.
channel_map : list_of_ints

Represents mapping from logical AI channel number to physical connector on the DAQ (varies between DAQ models). The default assumes DAQ 6363 and similar with AI channels on connectors 0-7 and 16-23.

default: (0, 1, 2, 3, 4, 5, 6, 7, 16, 17, 18, 19, 20, 21, 22, 23)

labels : list_of_strs
List of port labels. If specified, the lenght of the list must match the length of resistor_values. Defaults to “PORT_<pnum>”, where “pnum” is the number of the port.
negative_samples : str

Specifies how negative power samples should be handled. The following methods are possible:

keep:keep them as they are
zero:turn negative values to zero
drop:drop samples if they contain negative values. warning: this may result in port files containing different numbers of samples
abs:take the absoulte value of negave samples

allowed values: 'keep', 'zero', 'drop', 'abs'

default: 'keep'

gpio_sync : integer

If specified, the instrument will simultaneously set the specified GPIO pin high and put a marker into ftrace. This is to facillitate syncing kernel trace events to DAQ power trace.

constraint: value > 0

merge_channels : dict_or_bool

If set to True, channels with consecutive letter suffixes will be summed. e.g. If you have channels A7a, A7b, A7c, A15a, A15b they will be summed to A7, A15

You can also manually specify the name of channels to be merged and the name of the result like so:

merge_channels:
A15: [A15dvfs, A15ram] NonCPU: [GPU, RoS, Mem]

In the above exaples the DAQ channels labeled A15a and A15b will be summed together with the results being saved as ‘channel’ ‘’a’‘. A7, GPU and RoS will be summed to ‘c’

delay

This instrument introduces a delay before executing either an iteration or all iterations for a spec.

The delay may be specified as either a fixed period or a temperature threshold that must be reached.

Optionally, if an active cooling solution is employed to speed up temperature drop between runs, it may be controlled using this instrument.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
temperature_file : str

Full path to the sysfile on the device that contains the device’s temperature.

default: '/sys/devices/virtual/thermal/thermal_zone0/temp'

temperature_timeout : integer

The timeout after which the instrument will stop waiting even if the specified threshold temperature is not reached. If this timeout is hit, then a warning will be logged stating the actual temperature at which the timeout has ended.

default: 600

temperature_poll_period : integer

How long to sleep (in seconds) between polling current device temperature.

default: 5

temperature_between_specs : integer

Temperature (in device-specific units) the device must cool down to before the iteration spec will be run.

Note

This cannot be specified at the same time as fixed_between_specs

temperature_between_iterations : integer

Temperature (in device-specific units) the device must cool down to before the next spec will be run.

Note

This cannot be specified at the same time as fixed_between_iterations

temperature_before_start : integer

Temperature (in device-specific units) the device must cool down to just before the actual workload execution (after setup has been performed).

Note

This cannot be specified at the same time as fixed_between_iterations

fixed_between_specs : integer

How long to sleep (in seconds) after all iterations for a workload spec have executed.

Note

This cannot be specified at the same time as temperature_between_specs

fixed_between_iterations : integer

How long to sleep (in seconds) after each iterations for a workload spec has executed.

Note

This cannot be specified at the same time as temperature_between_iterations

fixed_before_start : integer

How long to sleep (in seconds) after setup for an iteration has been perfromed but before running the workload.

Note

This cannot be specified at the same time as temperature_before_start

active_cooling : boolean
This instrument supports an active cooling solution while waiting for the device temperature to drop to the threshold. The solution involves an mbed controlling a fan. The mbed is signaled over a serial port. If this solution is present in the setup, this should be set to True.

dmesg

Collected dmesg output before and during the run.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
loglevel : integer

Set loglevel for console output.

allowed values: 0, 1, 2, 3, 4, 5, 6, 7

energy_model

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
device_name : caseless_string
The name of the device to be used in generating the model. If not specified, device.name will be used.
big_core : caseless_string
The name of the “big” core in the big.LITTLE system; must match one of the values in device.core_names.
performance_metric : caseless_string (mandatory)
Metric to be used as the performance indicator.
power_metric : list_or_caseless_string
Metric to be used as the power indicator. The value may contain a {core} format specifier that will be replaced with names of big and little cores to drive the name of the metric for that cluster. Ether this or energy_metric must be specified but not both.
energy_metric : list_or_caseless_string
Metric to be used as the energy indicator. The value may contain a {core} format specifier that will be replaced with names of big and little cores to drive the name of the metric for that cluster. this metric will be used to derive power by deviding through by execution time. Either this or power_metric must be specified, but not both.
power_scaling_factor : float

Power model specfies power in milliWatts. This is a scaling factor that power_metric values will be multiplied by to get milliWatts.

default: 1.0

big_frequencies : list_of_ints
List of frequencies to be used for big cores. These frequencies must be supported by the cores. If this is not specified, all available frequencies for the core (as read from cpufreq) will be used.
little_frequencies : list_of_ints
List of frequencies to be used for little cores. These frequencies must be supported by the cores. If this is not specified, all available frequencies for the core (as read from cpufreq) will be used.
idle_workload : str

Workload to be used while measuring idle power.

default: 'idle'

idle_workload_params : dict
Parameter to pass to the idle workload.
first_cluster_idle_state : integer

The index of the first cluster idle state on the device. Previous states are assumed to be core idles. The default is -1, i.e. only the last idle state is assumed to affect the entire cluster.

default: -1

no_hotplug : boolean
This options allows running the instrument without hotpluging cores on and off. Disabling hotplugging will most likely produce a less accurate power model.
num_of_freqs_to_thermal_adjust : integer
The number of frequencies begining from the highest, to be adjusted for the thermal effect.
big_opps : opp_table
OPP table mapping frequency to voltage (kHz –> mV) for the big cluster.
little_opps : opp_table
OPP table mapping frequency to voltage (kHz –> mV) for the little cluster.
big_leakage : integer

Leakage factor for the big cluster (this is specific to a particular core implementation).

default: 120

little_leakage : integer

Leakage factor for the little cluster (this is specific to a particular core implementation).

default: 60

energy_probe

Collects power traces using the ARM energy probe.

This instrument requires caiman utility to be installed in the workload automation host and be in the PATH. Caiman is part of DS-5 and should be in /path/to/DS-5/bin/ . Energy probe can simultaneously collect energy from up to 3 power rails.

To connect the energy probe on a rail, connect the white wire to the pin that is closer to the Voltage source and the black wire to the pin that is closer to the load (the SoC or the device you are probing). Between the pins there should be a shunt resistor of known resistance in the range of 5 to 20 mOhm. The resistance of the shunt resistors is a mandatory parameter resistor_values.

Note

This instrument can process results a lot faster if python pandas is installed.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
resistor_values : list_of_numbers
The value of shunt resistors. This is a mandatory parameter.
labels : list
Meaningful labels for each of the monitored rails.
device_entry : str

Path to /dev entry for the energy probe (it should be /dev/ttyACMx)

default: '/dev/ttyACM0'

execution_time

Measure how long it took to execute the run() methods of a Workload.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.

file_poller

Polls the given files at a set sample interval. The values are output in CSV format.

This instrument places a file called poller.csv in each iterations result directory. This file will contain a timestamp column which will be in uS, the rest of the columns will be the contents of the polled files at that time.

This instrument will strip any commas or new lines for the files’ values before writing them.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
sample_interval : integer

The interval between samples in mS.

default: 1000

files : list_or_string (mandatory)
A list of paths to the files to be polled
labels : list_or_string
A list of lables to be used in the CSV output for the corresponding files. This cannot be used if a * wildcard is used in a path.
as_root : boolean
Whether or not the poller will be run as root. This should be used when the file you need to poll can only be accessed by root.

fps

Measures Frames Per Second (FPS) and associated metrics for a workload.

Note

This instrument depends on pandas Python library (which is not part of standard WA dependencies), so you will need to install that first, before you can use it.

Android L and below use SurfaceFlinger to calculate the FPS data. Android M and above use gfxinfo to calculate the FPS data.

SurfaceFlinger: The view is specified by the workload as view attribute. This defaults to 'SurfaceView' for game workloads, and None for non-game workloads (as for them FPS mesurement usually doesn’t make sense). Individual workloads may override this.

gfxinfo: The view is specified by the workload as package attribute. This is because gfxinfo already processes for all views in a package.

This instrument adds four metrics to the results:

FPS:Frames Per Second. This is the frame rate of the workload.
frame_count:The total number of frames rendered during the execution of the workload.
janks:The number of “janks” that occured during execution of the workload. Janks are sudden shifts in frame rate. They result in a “stuttery” UI. See http://jankfree.org/jank-busters-io
not_at_vsync:The number of frames that did not render in a single vsync cycle.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
drop_threshold : numeric

Data points below this FPS will be dropped as they do not constitute “real” gameplay. The assumption being that while actually running, the FPS in the game will not drop below X frames per second, except on loading screens, menus, etc, which should not contribute to FPS calculation.

default: 5

keep_raw : boolean
If set to True, this will keep the raw dumpsys output in the results directory (this is maily used for debugging) Note: frames.csv with collected frames data will always be generated regardless of this setting.
generate_csv : boolean

If set to True, this will produce temporal fps data in the results directory, in a file named fps.csv Note: fps data will appear as discrete step-like values in order to produce a more meainingfull representation,a rolling mean can be applied.

default: True

crash_check : boolean

Specifies wither the instrument should check for crashed content by examining frame data. If this is set, execution_time instrument must also be installed. The check is performed by using the measured FPS and exection time to estimate the expected frames cound and comparing that against the measured frames count. The the ratio of measured/expected is too low, then it is assumed that the content has crashed part way during the run. What is “too low” is determined by crash_threshold.

Note

This is not 100% fool-proof. If the crash occurs sufficiently close to workload’s termination, it may not be detected. If this is expected, the threshold may be adjusted up to compensate.

default: True

crash_threshold : float

Specifies the threshold used to decided whether a measured/expected frames ration indicates a content crash. E.g. a value of 0.75 means the number of actual frames counted is a quarter lower than expected, it will treated as a content crash.

default: 0.7

dumpsys_period : float

Specifies the time period between calls to dumpsys SurfaceFlinger --latency in seconds when collecting frame data. Using a lower value improves the granularity of timings when recording actions that take a short time to complete. Note, this will produce duplicate frame data in the raw dumpsys output, however, this is filtered out in frames.csv. It may also affect the overall load on the system.

The default value of 2 seconds corresponds with the NUM_FRAME_RECORDS in android/services/surfaceflinger/FrameTracker.h (as of the time of writing currently 128) and a frame rate of 60 fps that is applicable to most devices.

constraint: value > 0

default: 2

force_surfaceflinger : boolean
By default, the method to capture fps data is based on Android version. If this is set to true, force the instrument to use the SurfaceFlinger method regardless of its Android version.

freq_sweep

Sweeps workloads through all available frequencies on a device.

When enabled this instrument will take all workloads specified in an agenda and run them at all available frequencies for all clusters.

Recommendations:
  • Setting the runner to ‘by_spec’ increases the chance of successfully completing an agenda without encountering hotplug issues
  • If possible disable dynamic hotplug on the target device
  • This instrument does not automatically pin workloads to the cores being swept since it is not aware of what the workloads do. To achieve this use the workload’s taskset parameter (if it has one).

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
sweeps : list

By default this instrument will sweep across all available frequencies for all available clusters. If you wish to only sweep across certain frequencies on particular clusters you can do so by specifying this parameter.

Sweeps should be a lists of dictionaries that can contain:
  • Cluster (mandatory): The name of the cluster this sweep will be performed on. E.g A7
  • Frequencies: A list of frequencies (in KHz) to use. If this is not provided all frequencies available for this cluster will be used. E.g: [800000, 900000, 100000]
  • label: Workload specs will be named {spec id}_{label}_{frequency}. If a label is not provided it will be named sweep{sweep No.}

Example sweep specification:

freq_sweep:
    sweeps:
        - cluster: A53
          label: littles
          frequencies: [800000, 900000, 100000]
        - cluster: A57
          label: bigs

hwmon

Hardware Monitor (hwmon) is a generic Linux kernel subsystem, providing access to hardware monitoring components like temperature or voltage/current sensors.

The following web page has more information:

You can specify which sensors HwmonInstrument looks for by specifying hwmon_sensors in your config.py, e.g.

hwmon_sensors = ['energy', 'temp']

If this setting is not specified, it will look for all sensors it knows about. Current valid values are:

:energy: Collect energy measurements and report energy consumed
         during run execution (the diff of before and after readings)
         in Joules.
:temp: Collect temperature measurements and report the before and
       after readings in degrees Celsius.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
sensors : list_of_strs

The kinds of sensors hwmon instrument will look for

allowed values: 'energy', 'temp'

default: ['energy', 'temp']

interrupts

Pulls the /proc/interrupts file before and after workload execution and diffs them to show what interrupts occurred during that time.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.

juno_energy

Collects internal energy meter measurements from Juno development board.

This instrument was created because (at the time of creation) Juno’s energy meter measurements aren’t exposed through HWMON or similar standardized mechanism, necessitating a dedicated instrument to access them.

This instrument, and the readenergy executable it relies on are very much tied to the Juno platform and are not expected to work on other boards.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
period : float

Specifies the time, in Seconds, between polling energy counters.

default: 0.1

strict : boolean

Setting this to False will omit the check that the device is "juno". This is useful if the underlying board is actually Juno but WA connects via a different interface (e.g. generic_linux).

default: True

netstats

Measures transmit/receive network traffic on an Android divice on per-package basis.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
packages : list_of_strs
List of Android packages who’s traffic will be monitored. If unspecified, all packages in the device will be monitorred.
period : integer

Polling period for instrumentation on the device. Traffic statistics will be updated every period seconds.

default: 5

force_reinstall : boolean
If True, instrumentation APK will always be re-installed even if it already installed on the device.
uninstall_on_completion : boolean
If True, instrumentation will be uninstalled upon run completion.

perf

Perf is a Linux profiling with performance counters.

Performance counters are CPU hardware registers that count hardware events such as instructions executed, cache-misses suffered, or branches mispredicted. They form a basis for profiling applications to trace dynamic control flow and identify hotspots.

pref accepts options and events. If no option is given the default ‘-a’ is used. For events, the default events are migrations and cs. They both can be specified in the config file.

Events must be provided as a list that contains them and they will look like this

perf_events = ['migrations', 'cs']

Events can be obtained by typing the following in the command line on the device

perf list

Whereas options, they can be provided as a single string as following

perf_options = '-a -i'

Options can be obtained by running the following in the command line

man perf-record

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
events : list_of_strs

Specifies the events to be counted.

constraint: must not be empty.

default: ['migrations', 'cs']

optionstring : list_or_string

Specifies options to be used for the perf command. This may be a list of option strings, in which case, multiple instances of perf will be kicked off – one for each option string. This may be used to e.g. collected different events from different big.LITTLE clusters.

default: '-a'

labels : list_of_strs
Provides labels for pref output. If specified, the number of labels must match the number of optionstrings.
force_install : boolean
always install perf binary even if perf is already present on the device.

screenon

Ensure screen is on before each iteration on Android devices.

A very basic instrument that checks that the screen is on on android devices. Optionally, it call poll the device periodically to ensure that the screen is still on.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
polling_period : integer
Set this to a non-zero value to enable periodic (every polling_period seconds) polling of the screen on the device to ensure it is on during a run.

servo_power

Collects power traces using the Chromium OS Servo Board.

Servo is a debug board used for Chromium OS test and development. Among other uses, it allows access to the built in power monitors (if present) of a Chrome OS device. More information on Servo board can be found in the link bellow:

https://www.chromium.org/chromium-os/servo

In order to use this instrument you need to be a sudoer and you need a chroot environment. More information on the chroot environment can be found on the link bellow:

https://www.chromium.org/chromium-os/developer-guide

If you wish to run servod on a remote machine you will need to allow it to accept external connections using the –host command line option, like so: sudo servod -b some_board -c some_board.xml –host=’‘

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
power_domains : list_of_strs
The names of power domains to be monitored by the instrument using servod.
labels : list_of_strs
Meaningful labels for each of the monitored domains.
chroot_path : str
Path to chroot direcory on the host.
sampling_rate : integer

Samples per second.

default: 10

board_name : str (mandatory)
The name of the board under test.
autostart : boolean

Automatically start servod. Set to False if you want to use an already running servod instance or a remote servo

default: True

host : str

When autostart is set to False you can specify the host on which servod is running allowing you to remotelly access as servo board.

if autostart is True this parameter is ignored and localhost is used instead

default: 'localhost'

port : integer

When autostart is set to false you must provide the port that servod is running on

If autostart is True this parameter is ignored and the port output during the startup of servod will be used.

default: 9999

vid : str
When more than one servo is plugged in, you must provide a vid/pid pair to identify the servio you wish to use.
pid : str
When more than one servo is plugged in, you must provide a vid/pid pair to identify the servio you wish to use.

streamline

Collect Streamline traces from the device.

Note

This instrument supports streamline that comes with DS-5 5.17 and later earlier versions of streamline may not work correctly (or at all).

This Instrument allows collecting streamline traces (such as PMU counter values) from the device. It assumes you have DS-5 (which Streamline is part of) installed on your system, and that streamline command is somewhere in PATH.

Streamline works by connecting to gator service on the device. gator comes in two parts a driver (gator.ko) and daemon (gatord). The driver needs to be compiled against your kernel and both driver and daemon need to be compatible with your version of Streamline. The best way to ensure compatibility is to build them from source which came with your DS-5. gator source can be found in

/usr/local/DS-5/arm/gator

(the exact path may vary depending of where you have installed DS-5.) Please refer to the README the accompanies the source for instructions on how to build it.

Once you have built the driver and the daemon, place the binaries into your ~/.workload_automation/streamline/ directory (if you haven’t tried running WA with this instrument before, the streamline/ subdirectory might not exist, in which case you will need to create it.

In order to specify which events should be captured, you need to provide a configuration.xml for the gator. The easiest way to obtain this file is to export it from event configuration dialog in DS-5 streamline GUI. The file should be called “configuration.xml” and it be placed in the same directory as the gator binaries.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
port : str

Specifies the port on which streamline will connect to gator

default: '8080'

configxml : str
streamline configuration XML file to be used. This must be an absolute path, though it may count the user home symbol (~)
report : boolean
Specifies whether a report should be generated from streamline data.
report_options : str

A string with options that will be added to streamline -report command.

default: '-format csv'

sysfs_extractor

Collects the contest of a set of directories, before and after workload execution and diffs the result.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
paths : list_of_strs (mandatory)
A list of paths to be pulled from the device. These could be directories as well as files.
use_tmpfs : boolean
Specifies whether tmpfs should be used to cache sysfile trees and then pull them down as a tarball. This is significantly faster then just copying the directory trees from the device directly, bur requres root and may not work on all devices. Defaults to True if the device is rooted and False if it is not.
tmpfs_mount_point : str
Mount point for tmpfs partition used to store snapshots of paths.
tmpfs_size : str

Size of the tempfs partition.

default: '32m'

systrace

This instrument uses systrace.py from the android SDK to dump atrace output.

Note: This is unlikely to work on devices that have an android build built
before 15-May-2015. Before this date there was a bug with running atrace asynchronously.

From developer.android.com: The Systrace tool helps analyze the performance of your application by capturing and displaying execution times of your applications processes and other Android system processes. The tool combines data from the Android kernel such as the CPU scheduler, disk activity, and application threads to generate an HTML report that shows an overall picture of an Android device’s system processes for a given period of time.

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
buffer_size : integer

Use a trace buffer size of N kilobytes. This option lets you limit the total size of the data collected during a trace.

default: 1024

use_circular_buffer : boolean
When true trace data will be put into a circular buffer such that when it overflows it will start overwriting the beginning of the buffer.
kernel_functions : list_of_strs
Specify the names of kernel functions to trace.
categories : list_of_strs

A list of the categories you wish to trace.

default: ['freq', 'sched']

app_names : list_of_strs
Enable tracing for applications, specified as a comma-separated list of package names. The apps must contain tracing instrumentation calls from the Trace class. For more information, see http://developer.android.com/tools/debugging/systrace.html#app-trace
ignore_signals : boolean
This will cause atrace to ignore SIGHUP, SIGINT, SIGQUIT and SIGTERM.
compress_trace : boolean

Compresses atrace output. This greatly decreases the time it takes to pull results from a device but the resulting txt file is not human readable.

default: True

trace-cmd

trace-cmd is an instrument which interacts with ftrace Linux kernel internal tracer

From trace-cmd man page:

trace-cmd command interacts with the ftrace tracer that is built inside the Linux kernel. It interfaces with the ftrace specific files found in the debugfs file system under the tracing directory.

trace-cmd reads a list of events it will trace, which can be specified in the config file as follows

trace_events = ['irq*', 'power*']

If no event is specified, a default set of events that are generally considered useful for debugging/profiling purposes will be enabled.

The list of available events can be obtained by rooting and running the following command line on the device

trace-cmd list

You may also specify trace_buffer_size setting which must be an integer that will be used to set the ftrace buffer size. It will be interpreted as KB:

trace_cmd_buffer_size = 8000

The maximum buffer size varies from device to device, but there is a maximum and trying to set buffer size beyond that will fail. If you plan on collecting a lot of trace over long periods of time, the buffer size will not be enough and you will only get trace for the last portion of your run. To deal with this you can set the trace_mode setting to 'record' (the default is 'start'):

trace_cmd_mode = 'record'

This will cause trace-cmd to trace into file(s) on disk, rather than the buffer, and so the limit for the max size of the trace is set by the storage available on device. Bear in mind that 'record' mode is more intrusive than the default, so if you do not plan on generating a lot of trace, it is best to use the default 'start' mode.

Note

Mode names correspond to the underlying trace-cmd executable’s command used to implement them. You can find out more about what is happening in each case from trace-cmd documentation: https://lwn.net/Articles/341902/.

This instrument comes with an trace-cmd binary that will be copied and used on the device, however post-processing will be, by default, done on-host and you must have trace-cmd installed and in your path. On Ubuntu systems, this may be done with:

sudo apt-get install trace-cmd

Alternatively, you may set report_on_target parameter to True to enable on-target processing (this is useful when running on non-Linux hosts, but is likely to take longer and may fail on particularly resource-constrained targets).

parameters

modules : list
Lists the modules to be loaded by this extension. A module is a plug-in that further extends functionality of an extension.
events : list

Specifies the list of events to be traced. Each event in the list will be passed to trace-cmd with -e parameter and must be in the format accepted by trace-cmd.

default: ['sched*', 'irq*', 'power*']

mode : str

Trace can be collected using either ‘start’ or ‘record’ trace-cmd commands. In ‘start’ mode, trace will be collected into the ftrace buffer; in ‘record’ mode, trace will be written into a file on the device’s file system. ‘start’ mode is (in theory) less intrusive than ‘record’ mode, however it is limited by the size of the ftrace buffer (which is configurable – see buffer_size – but only up to a point) and that may overflow for long-running workloads, which will result in dropped events.

allowed values: 'start', 'record'

default: 'start'

buffer_size : integer
Attempt to set ftrace buffer size to the specified value (in KB). Default buffer size may need to be increased for long-running workloads, or if a large number of events have been enabled. Note: there is a maximum size that the buffer can be set, and that varies from device to device. Attempting to set buffer size higher than this will fail. In that case, this instrument will set the size to the highest possible value by going down from the specified size in buffer_size_step intervals.
buffer_size_step : integer

Defines the decremental step used if the specified buffer_size could not be set. This will be subtracted form the buffer size until set succeeds or size is reduced to 1MB.

default: 1000

buffer_size_file : str

Path to the debugs file that may be used to set ftrace buffer size. This should need to be modified for the vast majority devices.

default: '/sys/kernel/debug/tracing/buffer_size_kb'

report : boolean

Specifies whether reporting should be performed once the binary trace has been generated.

default: True

no_install : boolean
Do not install the bundled trace-cmd and use the one on the device instead. If there is not already a trace-cmd on the device, an error is raised.
report_on_target : boolean

When enabled generation of reports will be done host-side because the generated file is very large. If trace-cmd is not available on the host device this setting and be disabled and the report will be generated on the target device.

Note

This requires the latest version of trace-cmd to be installed on the host (the one in your distribution’s repos may be too old).