dgtimer

Go to: Synopsis. Return value. Keywords. Related. Flags. Python examples.

Synopsis

dgtimer([combineType=boolean], [hide=string], [hierarchy=boolean], [maxDisplay=int], [name=string], [noHeader=boolean], [outputFile=string], [overhead=boolean], [rangeLower=float], [rangeUpper=float], [reset=boolean], [returnCode=string], [returnType=string], [show=string], [sortMetric=string], [sortType=string], [threshold=float], [timerOff=boolean], [timerOn=boolean], [trace=boolean], [type=string], [uniqueName=boolean], [updateHeatMap=int])

Note: Strings representing object names and arguments must be separated by commas. This is not depicted in the synopsis.

dgtimer is NOT undoable, queryable, and NOT editable.

The various operations we measure are called "metrics" and the types of timers are called "timer types". The various metrics are always measured when timing is on, but are only queried when specified via the -show and -hide flags. The metrics currently supported are listed in detail under the -show flag below. For each metric we support a standard set of timer types. There are three of these: "self" for self time (the time specific to the node and not its children), "inclusive" (time including children of the node), and "count" (number of operations of the given metric on the node).

The timing mechanism which is used by dgtimer is built into the DG itself, thus ALL depend nodes can be timed and there is no need for programmers writing plug-ins using the OpenMaya API to add any special code in order for their nodes to be timed -- its all handled transparently.

The dgtimer command allows node timers to be turned on, off, reset to zero, and have their current value displayed, and these operations can be performed globally on all nodes or on a specific set of nodes defined by name, type or parentage. Note that all timer measurements are computed in "real time" (the same time measurement you get from a wristwatch) as opposed to "CPU time" (which only measures time when the processor is executing your code). All times are displayed in seconds.

Use the -query flag to display the current timer values on a node,
use -on to turn on timing,
use -off to turn off timing,
and -reset to reset timers to zero.

To display the values measured during timing, there are two approaches. The first method is to use the -query flag can be used to report the information which has been measured. The second method is to use the query methods available on the OpenMaya class MFnDependencyNode (see the OpenMaya documentation for details).

What follows is a description of what is generated via -query. The output is broken out into several sections and these are described as follows:

SECTION 1:
Section 1 of the dgtimer output contains global information. This section can be disabled via the -hoHeader flag. These values are reset whenever a global timer reset occurs (i.e. dgtimer -reset; is specified). The global values which are reported are:

• Total real time: the total wall-clock time since the last global timer reset. This is the actual time which has been spent as you might measure it measure it with your watch. On a multi-processing system, this value will always remain true to to real time (unlike user and sys time).
• Total user time: the total time the CPU(s) spent processing Maya not including any system time since the last global timer reset.
• Total sys time: the total time the CPU(s) spent in operating system calls on behalf of Maya since the last global timer reset.
• Summary of each metric for all nodes: a summary of self and count for each metric that we measure:
• Real time in callbacks reports the self time and count for the "callback" metric.
• Real time in compute reports the self time and count for the "compute" metric.
• Real time in dirty propagation reports the self time and count for the "dirty" metric.
• Real time in drawing reports the self time and count for the "draw" metric.
• Real time fetching data from plugs reports the self time and count for the "fetch" metric.
• Breakdown of select metrics in greater detail: a reporting of certain combinations of metrics that we measure:
• Real time in compute invoked from callback reports the self time spent in compute when invoked either directly or indirectly by a callback.
• Real time in compute not invoked from callback reports the self time spent in compute not invoked either directly or indirectly by a callback.

The columns which are displayed are as follows:

• Rank: The order of this node in the sorted list of all nodes, where the list is sorted by -sortMetric and -sortType flag values (if these are omitted the default is to sort by self compute time).
• ON: Tells you if the timer for that node is currently on or off. (With dgtimer, you have the ability to turn timing on and off on a per-node basis).
• Per-metric information: various columns are reported for each metric. The name of the metric is reported at in the header in capital letters (e.g. "DRAW"). The standard columns for each metric are:
• Self: The amount of real time (i.e. elapsed time as you might measure it with a stopwatch) spent performing the operation (thus if the metric is "DRAW", then this will be time spent drawing the node).
• Inclusive: The amount of real time (i.e. elapsed time as you might measure it with a stopwatch) spent performing the operation including any child operations that were invoked on behalf of the operation (thus if the metric is "DRAW", then this will be the total time taken to draw the node including any child operations).
• Count: The number of operations that occued on this node (thus if the metric is "DRAW", then the number of draw operations on the node will be reported).
• Sort information if a column is the one being used to sort all the per-node dgtimer information, then that column is followed by a Percent and Cumulative column which describe a running total through the listing. Note that "-sortType none" prevents these two columns from appearing and implicitely sorts on "self" time.
• After the per-metric columns, the node name and type are reported:
• Type The node type.
• Name The name of the node. If the node is file referenced and you are using namespaces, the namespace will be included. You can also force the dagpath to be displayed by specifying the -uniqueName flag.
• Plug-in name If the node was implemented in an OpenMaya plug-in, the name of that plug-in is reported.

The first part is SECTION 3.1 lists the time per callback with each entry comprising:

• The name of the callback, such as "attributeChangedMsg". These names are internal Maya names, and in the cases where the callback is available through the OpenMaya API, the API access to the callback is similarly named.
• The name is followed by a breakdown per callbackId. The callbackId is an identifying number which is unique to each client that is registered to a callback and can be deduced by the user, such as through the OpenMaya API. You can cross-reference by finding the same callbackId value listed in SECTIONs 3.1 and 3.3.
• Self time (i.e. real time spent within that callbackId type not including any child operations which occur while processing the callback).
• Percent (see the -sortType flag). Note that the percent values are listed to sum up to 100% for that callback. This is not a global percent.
• Cumulative (see the -sortType flag).
• Inclusive time (i.e. real time spent within that callbackId including any child operations).
• Count (number of times the callbackId was invoked).
• API lists "Y" if the callbackId was defined through the OpenMaya API, and "N" if the callbackId was defined internally within Maya.
• Node lists the name of the node this callbackId was associated with. If the callbackId was associated with more than one node, the string "*multiple*" is printed. If there was no node associated with the callbackId (or its a callback type in which the node is hard to deduce), the entry is blank.
• After the callbackId entries are listed, a dashed line is printed followed by a single line listing the self, inclusive and count values for the callback. Note that the percent is relative to the global callback time.

The second part is SECTION 3.2 which lists the data per callbackId. As noted earlier, the callbackId is an identifying number which is unique to each client that is registered to a callback and can be deduced by the user, such as through the OpenMaya API. The entries in SECTION 3.2 appear as follows:

• CallbackId the numeric identifier for the callback. You can cross reference by finding the same callbackId value listed in SECTIONs 3.1 and 3.3.
• For each callbackId, the data is broken down per-callback:
• Callback the name of the callback, e.g. "attributeChangedMsg".
• Percent, Cumulative, Inclusive, Count, API and Node entries as described in SECTION 3.1.
• After the callback entries are listed for the callbackId, a dashed followed by a summary line is printed. The summary line lists the self, inclusive and count values for the callback. Note that the percent is relative to the global callback time.

The third part is SECTION 3.3 which lists data per-callback per-node. The nodes are sorted based on the -sortType flag, and for each node, the callbacks are listed, also sorted based on the -sortType flag. As this listing can be long, zero entries are not displayed. An important note for SECTION 3.3 is that only nodes which still exist are displayed. If a node has been deleted, no infromation is listed.

Return value

In query mode, return type is based on queried flag.

Keywords

Related

Flags

Long name (short name)	Argument types	Properties
combineType(ct)	boolean
Causes all nodes of the same type (e.g. animCurveTA) to be combined in the output display.
hide(hi)	string
This flag is the converse of -show. As with -show, it is a query-only flag which can be specified multiple times. If you do specify -hide, we display all columns except those listed by the -hide flags.
hierarchy(h)	boolean
Used to specify that a hierarchy of the dependency graph be affected, thus "-reset -hierarchy -name ball" will reset the timers on the node named "ball" and all of its descendents in the dependency graph.
maxDisplay(m)	int
Truncates the display so that only the most expenive "n" entries are printed in the output display.
name(n)	string
Used in conjunction with -reset or -query to specify the name of the node to reset or print timer values for. When querying a single timer, only a single line of output is generated (i.e. the global timers and header information is omitted). Note that one can force output to the script editor window via the "-outputFile MEL" option to make it easy to grab the values in a MEL script. Note: the -name and -type flag cannot be used together.
noHeader(nh)	boolean
Used in conjunction with -query to prevent any header or footer information from being printed. All that will be output is the per-node timing data. This option makes it easier to parse the output such as when you output the query to a file on disk using the -outputFile option.
outputFile(o)	string
Specifies where the output of timing or tracing is displayed. The flag takes a string argument which accepts three possible values: The name of a file on disk. Or the keyword "stdout", which causes output to be displayed on the terminal window (Linux and Macintosh), and the status window on Windows. Or the keyword "MEL", which causes output to be displayed in the Maya Script Editor (only supported with -query). The "stdout" setting is the default behaviour. This flag can be used with the -query flag as well as the -trace flag. When used with the -trace flag, any tracing output will be displayed on the destination specified by the -outputFile (or stdout if -outputFile is omitted). Tracing operations will continue to output to the destination until you specify the -trace and -outputFile flags again. When used with the -query flag, timing output will be displayed to the destination specified by the -outputFile (or "stdout" if -outputFile is omitted). Here are some examples of how to use the -query, -trace and -outputFile flags: Example: output the timing information to a single file on disk: dgtimer -on; // Turn on timing create some animated scene content; play -wait; // Play the scene through dgtimer -query -outputFile "/tmp/timing.txt" // Output node timing information to a file on disk Example: output the tracing information to a single file on disk: dgtimer -on; // Turn on timing create some animated scene content; dgtimer -trace on -outputFile "/tmp/trace.txt" // Turn on tracing and output the results to file play -wait; // Play the scene through; trace info goes to /tmp/trace.txt dgtimer -query; // But the timing info goes to the terminal window play -wait; // Play the scene again, trace info still goes to /tmp/trace.txt Example: two runs, outputting the trace information and timing information to separate files: dgtimer -on; // Turn on timing create some animated scene content; dgtimer -trace on -outputFile "/tmp/trace1.txt" // Turn on tracing and output the results to file play -wait; // Play the scene through dgtimer -query -outputFile "/tmp/query1.txt" // Output node timing information to another file dgtimer -reset; dgtimer -trace on -outputFile "/tmp/trace2.txt" // Output tracing results to different file play -wait; // Play the scene through dgtimer -query -outputFile "/tmp/query2.txt" // Output node timing information to another file Tips and tricks: Outputting the timing results to the script editor makes it easy to use the results in MEL e.g. string $timing[] = `dgtimer -query -outputFile MEL`. It is important to note that the -outputFile you specify with -trace is totally independent from the one you specify with -query. If the file you specify already exists, Maya will empty the file first before outputting data to it (and if the file is not writable, an error is generated instead). In query mode, this flag needs a value.
overhead(oh)	boolean
Turns on and off the measurement of timing overhead. Under ordinary circumstances the amount of timing overhead is minimal compared with the events being measured, but in complex scenes, one might find the overhead to be measurable. By default this option is turned off. To enable it, specify "dgtimer -overhead true" prior to starting timing. When querying timing, the overhead is reported in SECTION 1.2 of the dgtimer output and is not factored out of each individual operation.
rangeLower(rgl)	float
This flag can be specified to limit the range of nodes which are displayed in a query, or the limits of the heat map with -updateHeatMap. The value is the lower percentage cutoff for the nodes which are processed. There is also a -rangeLower flag which sets the lower range limit. The default value is 0, meaning that all nodes with timing value below the upper range limit are considered.
rangeUpper(rgu)	float
This flag can be specified to limit the range of nodes which are displayed in a query, or the limits of the heat map with -updateHeatMap. The value is the upper percentage cutoff for the nodes which are processed. There is also a -rangeLower flag which sets the lower range limit. The default value is 100, meaning that all nodes with timing value above the lower range limit are considered.
reset(r)	boolean
Resets the node timers to zero. By default, the timers on all nodes as well as the global timers are reset, but if specified with the -name or -type flags, only the timers on specified nodes are reset.
returnCode(rc)	string
This flag has been replaced by the more general -returnType flag. The -returnCode flag was unfortunately specific to the compute metric only. It exists only for backwards compatability purposes. It will be removed altogether in a future release. Here are some handy equivalences: To get the total number of nodes: OLD WAY: dgtimer -rc nodecount -q; // Result:325// NEW WAY: dgtimer -returnType total -sortType none -q; // Result:325// OLD WAY: dgtimer -rc count -q; // Result:1270// To get the sum of the compute count column: NEW WAY: dgtimer -returnType total -sortMetric compute -sortType count -q; // Result:1270// OLD WAY: dgtimer -rc selftime -q; // Result:0.112898// To get the sum of the compute self column: NEW WAY: dgtimer -returnType total -sortMetric compute -sortType self -q; // Result:0.112898//
returnType(rt)	string
This flag specifies what the double value returned by the dgtimer command represents. By default, the value returned is the global total as displayed in SECTION 1 for the column we are sorting on in the per-node output (the sort column can be specified via the -sortMetric and -sortType flags). However, instead of the total being returned, the user can instead request the individual entries for the column. This flag is useful mainly for querying without forcing any output. The flag accepts the values "total", to just display the column total, or "all" to display all entries individually. For example, if you want to get the total of draw self time without any other output simply specify the following: dgtimer -returnType total -sortMetric draw -sortType self -threshold 100 -noHeader -query; // Result: 7718.01 // To instead get each individual entry, change the above query to: dgtimer -returnType all -sortMetric draw -sortType self -threshold 100 -noHeader -query; // Result: 6576.01 21.91 11.17 1108.92 // To get the inclusive dirty time for a specific node, use -name as well as -returnType all: dgtimer -name "virginia" -returnType all -sortMetric dirty -sortType inclusive -threshold 100 -noHeader -query; Note: to get the total number of nodes, use "-sortType none -returnType total". To get the on/off status for each node, use "-sortType none -returnType all".
show(sh)	string
Used in conjunction with -query to specify which columns are to be displayed in the per-node section of the output. -show takes an argument, which can be "all" (to display all columns), "callback" (to display the time spent during any callback processing on the node not due to evaluation), "compute" (to display the time spent in the node's compute methods), "dirty" (to display time spent propagating dirtiness on behalf of the node), "draw" (to display time spent drawing the node), "compcb" (to display time spent during callback processing on node due to compute), and "compncb" (to display time spent during callback processing on node NOT due to compute). The -show flag can be used multiple times, but cannot be specified with -hide. By default, if neither -show, -hide, or -sort are given, the effective display mode is: "dgtimer -show compute -query".
sortMetric(sm)	string
Used in conjunction with -query to specify which metric is to be sorted on when the per-node section of the output is generated, for example "draw" time. Note that the -sortType flag can also be specified to define which timer is sorted on: for example "dgtimer -sortMetric draw -sortType count -query" will sort the output by the number of times each node was drawn. Both -sortMetric and -sortType are optional and you can specify one without the other. The -sortMetric flag can only be specified at most once. The flag takes the following arguments: "callback" (to sort on time spent during any callback processing on the node), "compute" (to sort on the time spent in the node's compute methods), "dirty" (to sort on the time spent propagating dirtiness on behalf of the node), "draw" (to sort on time spent drawing the node), "fetch" (to sort on time spent copying data from the datablock), The default, if -sortMetric is omitted, is to sort on the first displayed column. Note that the sortMetric is independent of which columns are displayed via -show and -hide. Sort on a hidden column is allowed. The column selected by -sortMetric and -sortType specifies which total is returned by the dgtimer command on the MEL command line. This flag is also used with -updateHeatMap to specify which metric to build the heat map for.
sortType(st)	string
Used in conjunction with -query to specify which timer is to be sorted on when the per-node section of the output is generated, for example "self" time. Note that the -sortMetric flag can also be specified to define which metric is sorted on: for example "dgtimer -sortMetric draw -sortType count -query" will sort the output by the number of times each node was drawn. Both -sortMetric and -sortType are optional and you can specify one without the other. The -sortType flag can be specified at most once. The flag takes the following arguments: "self" (to sort on self time, which is the time specific to the node and not its children), "inclusive" (to sort on the time including children of the node), "count" (to sort on the number of times the node was invoked). and "none" (to sort on self time, but do not display the Percent and Cumulative columns in the per-node display, as well as cause the total number of nodes in Maya to be returned on the command line). The default, if -sortType is omitted, is to sort on self time. The column selected by -sortMetric and -sortType specifies which total is returned by the dgtimer command on the MEL command line. The global total as displayed in SECTION 1 of the listing is returned. The special case of "-sortType none" causes the number of nodes in Maya to instead be returned. This flag is also used with -updateHeatMap to specify which metric to build the heat map for.
threshold(th)	float
Truncates the display once the value falls below the threshold value. The threshold applies to whatever timer is being used for sorting. For example, if our sort key is self compute time (i.e. -sortMetric is "compute" and -sortType is "self") and the threshold parameter is 20.0, then only nodes with a compute self-time of 20.0 or higher will be displayed. (Note that -threshold uses absolute time. There are the similar -rangeUpper and -rangeLower parameters which specify a range using percentage).
timerOff(off)	boolean
Turns off node timing. By default, the timers on all nodes are turned off, but if specified with the -name or -type flags, only the timers on specified nodes are turned off. If the timers on all nodes become turned off, then global timing is also turned off as well.
timerOn(on)	boolean
Turns on node timing. By default, the timers on all nodes are turned on, but if specified with the -name or -type flags, only the timers on specified nodes are turned on. The global timers are also turned on by this command. Note that turning on timing does NOT reset the timers to zero. Use the -reset flag to reset the timers. The idea for NOT resetting the timers is to allow the user to arbitrarily turn timing on and off and continue to add to the existing timer values.
trace(tr)	boolean
Turns on or off detailed execution tracing. By default, tracing is off. If enabled, each timeable operation is logged when it starts and again when it ends. This flag can be used in conjunction with -outputFile to specify where the output is generated to. The following example shows how the output is formatted: dgtimer:begin: compute 3 particleShape1Deformed particleShape1Deformed.lastPosition The above is an example of the output when -trace is true that marks the start of an operation. For specific details on each field: the "dgtimer:begin:" string is an identifying marker to flag that this is a begin operation record. The second argument, "compute" in our example, is the operation metric. You can view the output of each given metric via "dgtimer -q" by specifying the -show flag. The integer which follows (3 in this case) is the depth in the operation stack, and the third argument is the name of the node (particleShape1Deformed). The fourth argument is specific to the metric. For "compute", it gives the name of the plug being computed. For "callback", its the internal Maya name of the callback. For "dirty", its the name of the plug that dirtiness is being propagated from. dgtimer:end: compute 3 particleShape1Deformed 0.000305685 0.000305685 The above is the end operation record. The "compute", "3" and "particleShapeDeformed" arguments were described in the "dgtimer:begin" overview earlier. The two floating-point arguments are self time and inclusive time for the operation measured in seconds. The inclusive measure lists the total time since the matching "dgtimer:begin:" entry for this operation, while the self measure lists the inclusive time minus any time consumed by child operations which may have occurred during execution of the current operation. As noted elsewhere in this document, these two times are "wall clock times", measuring elapsed time including any time in which Maya was idle or performing system calls. Since dgtimer can measure some non-node qualities in Maya, such as global message callbacks, a "-" is displayed where the node name would ordinarily be displayed. The "-" means "not applicable".
type(t)	string
Used in conjunction with -reset or -query to specify the type of the node(s) (e.g. animCurveTA) to reset or print timer values for. When querying, use of the -combineType flag will cause all nodes of the same type to be combined into one entry, and only one line of output is generated (i.e. the global timers and header information is omitted). Note: the -name and -type flag cannot be used together.
uniqueName(un)	boolean
Used to specify that the DAG nodes listed in the output should be listed by their unique names. The full DAG path to the object will be printed out instead of just the node name.
updateHeatMap(uhm)	int
Forces Maya's heat map to rebuild based on the specified parameters. The heat map is an internal dgtimer structure used in mapping intensity values to colourmap entries during display by the HyperGraph Editor. There is one heat map shared by all editors that are using heat map display mode. Updating the heat map causes the timer values on all nodes to be analysed to compose the distribution of entries in the heat map. The parameter is the integer number of divisions in the map and should equal the number of available colours for displaying the heat map. This flag can be specified with the -rangeLower and -rangeUpper flags to limit the range of displayable to lie between the percentile range. The dgtimer command returns the maximum timing value for all nodes in Maya for the specified metric and type. Note: when the display range includes 0, the special zeroth (exactly zero) slot in the heat map is avilable.

Flag can appear in Create mode of command	Flag can appear in Edit mode of command
Flag can appear in Query mode of command	Flag can have multiple arguments, passed either as a tuple or a list.

Python examples

import maya.cmds as cmds

# Turns on node timing and resets the timers.
cmds.dgtimer( on=True )
# Turns off node timing. Note that this does not reset the
# timers.
cmds.dgtimer( off=True )
# Prints the current timer values to the default (stdout).
cmds.dgtimer( query=True )
# To reset the timers:
cmds.dgtimer( reset=True )
# Turn on node timing and reset the timer values to zero.
# Then, playback the scene, turn off timing and dump to a file.
# Turn on timing without resetting the timers, and repeat.
cmds.dgtimer( on=True, reset=True )
cmds.play( wait=True )
cmds.dgtimer( off=True )
cmds.dgtimer( outputFile='/home/virginia/timing/dgtrace_once.txt', query=True )
cmds.dgtimer( on=True )
cmds.play( wait=True )
cmds.dgtimer( off=True )
cmds.dgtimer( outputFile='/home/virginia/timing/dgtrace_twice.txt', query=True )