starpu-max/doc/doxygen/chapters/380_offline_performance_tools.doxy

/*
 * This file is part of the StarPU Handbook.
 * Copyright (C) 2009--2011  Universit@'e de Bordeaux
 * Copyright (C) 2010, 2011, 2012, 2013, 2014, 2016  CNRS
 * Copyright (C) 2011, 2012 INRIA
 * See the file version.doxy for copying conditions.
 */

/*! \page OfflinePerformanceTools Offline Performance Tools

To get an idea of what is happening, a lot of performance feedback is available,
detailed in this chapter. The various informations should be checked for.

<ul>
<li>
What does the Gantt diagram look like? (see \ref CreatingAGanttDiagram)
<ul>
  <li> If it's mostly green (tasks running in the initial context) or context specific
  color prevailing, then the machine is properly
  utilized, and perhaps the codelets are just slow. Check their performance, see
  \ref PerformanceOfCodelets.
  </li>
  <li> If it's mostly purple (FetchingInput), tasks keep waiting for data
  transfers, do you perhaps have far more communication than computation? Did
  you properly use CUDA streams to make sure communication can be
  overlapped? Did you use data-locality aware schedulers to avoid transfers as
  much as possible?
  </li>
  <li> If it's mostly red (Blocked), tasks keep waiting for dependencies,
  do you have enough parallelism? It might be a good idea to check what the DAG
  looks like (see \ref CreatingADAGWithGraphviz).
  </li>
  <li> If only some workers are completely red (Blocked), for some reason the
  scheduler didn't assign tasks to them. Perhaps the performance model is bogus,
  check it (see \ref PerformanceOfCodelets). Do all your codelets have a
  performance model?  When some of them don't, the schedulers switches to a
  greedy algorithm which thus performs badly.
  </li>
</ul>
</li>
</ul>

You can also use the Temanejo task debugger (see \ref UsingTheTemanejoTaskDebugger) to
visualize the task graph more easily.
\section Off-linePerformanceFeedback Off-line Performance Feedback

\subsection GeneratingTracesWithFxT Generating Traces With FxT

StarPU can use the FxT library (see
https://savannah.nongnu.org/projects/fkt/) to generate traces
with a limited runtime overhead.

You can either get a tarball:

\verbatim
$ wget http://download.savannah.gnu.org/releases/fkt/fxt-0.2.11.tar.gz
\endverbatim

or use the FxT library from CVS (autotools are required):

\verbatim
$ cvs -d :pserver:anonymous\@cvs.sv.gnu.org:/sources/fkt co FxT
$ ./bootstrap
\endverbatim

Compiling and installing the FxT library in the <c>$FXTDIR</c> path is
done following the standard procedure:

\verbatim
$ ./configure --prefix=$FXTDIR
$ make
$ make install
\endverbatim

In order to have StarPU to generate traces, StarPU should be configured with
the option \ref with-fxt "--with-fxt" :

\verbatim
$ ./configure --with-fxt=$FXTDIR
\endverbatim

Or you can simply point the <c>PKG_CONFIG_PATH</c> to
<c>$FXTDIR/lib/pkgconfig</c> and pass
\ref with-fxt "--with-fxt" to <c>./configure</c>

When FxT is enabled, a trace is generated when StarPU is terminated by calling
starpu_shutdown(). The trace is a binary file whose name has the form
<c>prof_file_XXX_YYY</c> where <c>XXX</c> is the user name, and
<c>YYY</c> is the pid of the process that used StarPU. This file is saved in the
<c>/tmp/</c> directory by default, or by the directory specified by
the environment variable \ref STARPU_FXT_PREFIX.

The additional configure option \ref enable-fxt-lock "--enable-fxt-lock" can
be used to generate trace events which describes the locks behaviour during
the execution.

\subsection CreatingAGanttDiagram Creating a Gantt Diagram

When the FxT trace file <c>prof_file_something</c> has been generated, it is possible to
generate a trace in the Paje format by calling:

\verbatim
$ starpu_fxt_tool -i /tmp/prof_file_something
\endverbatim

Or alternatively, setting the environment variable \ref STARPU_GENERATE_TRACE
to <c>1</c> before application execution will make StarPU do it automatically at
application shutdown.

This will create a file <c>paje.trace</c> in the current directory that
can be inspected with the ViTE (http://vite.gforge.inria.fr/) trace
visualizing open-source tool.  It is possible to open the
file <c>paje.trace</c> with ViTE by using the following command:

\verbatim
$ vite paje.trace
\endverbatim

To get names of tasks instead of "unknown", fill the optional
starpu_codelet::name, or use a performance model for them.
Details of the codelet execution can be obtained by passing
\ref enable-paje-codelet-details "--enable-paje-codelet-details" when
configuring StarPU and using a recent enough version of ViTE (at least
r1430).

In the MPI execution case, \ref STARPU_GENERATE_TRACE will not work as expected
(each node will try to generate paje.trace, thus mixing outputs...), you have to
collect the trace files from the MPI nodes, and
specify them all on the command <c>starpu_fxt_tool</c>, for instance:

\verbatim
$ starpu_fxt_tool -i /tmp/prof_file_something1 -i /tmp/prof_file_something2
\endverbatim

By default, all tasks are displayed using a green color. To display tasks with
varying colors, pass option <c>-c</c> to <c>starpu_fxt_tool</c>.

To identify tasks precisely, the application can set the starpu_task::tag_id field of the
task (or use \ref STARPU_TAG_ONLY when using starpu_task_insert()), and with a recent
enough version of ViTE (>= r1430) and the
\ref enable-paje-codelet-details "--enable-paje-codelet-details"
StarPU configure option, the value of the tag will show up in the trace.

It can also set the starpu_task::name field of the task (or use \ref STARPU_NAME)
when using starpu_task_insert()), to replace in traces the name of the codelet
with an arbitrarily chosen name.

Traces can also be inspected by hand by using the tool <c>fxt_print</c>, for instance:

\verbatim
$ fxt_print -o -f /tmp/prof_file_something
\endverbatim

Timings are in nanoseconds (while timings as seen in ViTE are in milliseconds).

\subsection CreatingADAGWithGraphviz Creating a DAG With Graphviz

When the FxT trace file <c>prof_file_something</c> has been generated, it is possible to
generate a task graph in the DOT format by calling:

\verbatim
$ starpu_fxt_tool -i /tmp/prof_file_something
\endverbatim

This will create a <c>dag.dot</c> file in the current directory. This file is a
task graph described using the DOT language. It is possible to get a
graphical output of the graph by using the graphviz library:

\verbatim
$ dot -Tpdf dag.dot -o output.pdf
\endverbatim

\subsection TraceTaskDetails Getting Task Details

When the FxT trace file <c>prof_file_something</c> has been generated, details on the
executed tasks can be retrieved by calling:

\verbatim
$ starpu_fxt_tool -i /tmp/prof_file_something
\endverbatim

This will create a <c>tasks.rec</c> file in the current directory.  This file
is in the recutils format, i.e. <c>Field: value</c> lines, and empty lines to
separate each task.  This can be used as a convenient input for various ad-hoc
analysis tools.  The performance models can be opened for instance by using
starpu_perfmodel_load_symbol() and then using
starpu_perfmodel_history_based_expected_perf().

\subsection MonitoringActivity Monitoring Activity

When the FxT trace file <c>prof_file_something</c> has been generated, it is possible to
generate an activity trace by calling:

\verbatim
$ starpu_fxt_tool -i /tmp/prof_file_something
\endverbatim

This will create a file <c>activity.data</c> in the current
directory. A profile of the application showing the activity of StarPU
during the execution of the program can be generated:

\verbatim
$ starpu_workers_activity activity.data
\endverbatim

This will create a file named <c>activity.eps</c> in the current directory.
This picture is composed of two parts.
The first part shows the activity of the different workers. The green sections
indicate which proportion of the time was spent executed kernels on the
processing unit. The red sections indicate the proportion of time spent in
StartPU: an important overhead may indicate that the granularity may be too
low, and that bigger tasks may be appropriate to use the processing unit more
efficiently. The black sections indicate that the processing unit was blocked
because there was no task to process: this may indicate a lack of parallelism
which may be alleviated by creating more tasks when it is possible.

The second part of the picture <c>activity.eps</c> is a graph showing the
evolution of the number of tasks available in the system during the execution.
Ready tasks are shown in black, and tasks that are submitted but not
schedulable yet are shown in grey.

\subsection Animation Getting Modular Schedular Animation

When using modular schedulers (i.e. schedulers which use a modular architecture,
and whose name start with "modular-"), the command

\verbatim
$ starpu_fxt_tool -i /tmp/prof_file_something
\endverbatim

will also produce a <c>trace.html</c> file which can be viewed in a
javascript-enabled web browser. It shows the flow of tasks between the
components of the modular scheduler.

\subsection LimitingScopeTrace Limiting The Scope Of The Trace

For computing statistics, it is useful to limit the trace to a given portion of
the time of the whole execution. This can be achieved by calling

\code{.c}
starpu_fxt_autostart_profiling(0)
\endcode

before calling starpu_init(), to
prevent tracing from starting immediately. Then

\code{.c}
starpu_fxt_start_profiling();
\endcode

and

\code{.c}
starpu_fxt_stop_profiling();
\endcode

can be used around the portion of code to be traced. This will show up as marks
in the trace, and states of workers will only show up for that portion.

\section PerformanceOfCodelets Performance Of Codelets

The performance model of codelets (see \ref PerformanceModelExample)
can be examined by using the tool <c>starpu_perfmodel_display</c>:

\verbatim
$ starpu_perfmodel_display -l
file: <malloc_pinned.hannibal>
file: <starpu_slu_lu_model_21.hannibal>
file: <starpu_slu_lu_model_11.hannibal>
file: <starpu_slu_lu_model_22.hannibal>
file: <starpu_slu_lu_model_12.hannibal>
\endverbatim

Here, the codelets of the example <c>lu</c> are available. We can examine the
performance of the kernel <c>22</c> (in micro-seconds), which is history-based:

\verbatim
$ starpu_perfmodel_display -s starpu_slu_lu_model_22
performance model for cpu
# hash      size       mean          dev           n
57618ab0    19660800   2.851069e+05  1.829369e+04  109
performance model for cuda_0
# hash      size       mean          dev           n
57618ab0    19660800   1.164144e+04  1.556094e+01  315
performance model for cuda_1
# hash      size       mean          dev           n
57618ab0    19660800   1.164271e+04  1.330628e+01  360
performance model for cuda_2
# hash      size       mean          dev           n
57618ab0    19660800   1.166730e+04  3.390395e+02  456
\endverbatim

We can see that for the given size, over a sample of a few hundreds of
execution, the GPUs are about 20 times faster than the CPUs (numbers are in
us). The standard deviation is extremely low for the GPUs, and less than 10% for
CPUs.

This tool can also be used for regression-based performance models. It will then
display the regression formula, and in the case of non-linear regression, the
same performance log as for history-based performance models:

\verbatim
$ starpu_perfmodel_display -s non_linear_memset_regression_based
performance model for cpu_impl_0
	Regression : #sample = 1400
	Linear: y = alpha size ^ beta
		alpha = 1.335973e-03
		beta = 8.024020e-01
	Non-Linear: y = a size ^b + c
		a = 5.429195e-04
		b = 8.654899e-01
		c = 9.009313e-01
# hash		size		mean		stddev		n
a3d3725e	4096           	4.763200e+00   	7.650928e-01   	100
870a30aa	8192           	1.827970e+00   	2.037181e-01   	100
48e988e9	16384          	2.652800e+00   	1.876459e-01   	100
961e65d2	32768          	4.255530e+00   	3.518025e-01   	100
...
\endverbatim

The same can also be achieved by using StarPU's library API, see
\ref API_Performance_Model and notably the function
starpu_perfmodel_load_symbol(). The source code of the tool
<c>starpu_perfmodel_display</c> can be a useful example.

The tool <c>starpu_perfmodel_plot</c> can be used to draw performance
models. It writes a <c>.gp</c> file in the current directory, to be
run with the tool <c>gnuplot</c>, which shows the corresponding curve.

\image html starpu_non_linear_memset_regression_based.png
\image latex starpu_non_linear_memset_regression_based.eps "" width=\textwidth

When the field starpu_task::flops is set (or \ref STARPU_FLOPS is passed to
starpu_task_insert()), <c>starpu_perfmodel_plot</c> can directly draw a GFlops
curve, by simply adding the <c>-f</c> option:

\verbatim
$ starpu_perfmodel_plot -f -s chol_model_11
\endverbatim

This will however disable displaying the regression model, for which we can not
compute GFlops.

\image html starpu_chol_model_11_type.png
\image latex starpu_chol_model_11_type.eps "" width=\textwidth

When the FxT trace file <c>prof_file_something</c> has been generated, it is possible to
get a profiling of each codelet by calling:

\verbatim
$ starpu_fxt_tool -i /tmp/prof_file_something
$ starpu_codelet_profile distrib.data codelet_name
\endverbatim

This will create profiling data files, and a <c>distrib.data.gp</c> file in the current
directory, which draws the distribution of codelet time over the application
execution, according to data input size.

\image html distrib_data.png
\image latex distrib_data.eps "" width=\textwidth

This is also available in the tool <c>starpu_perfmodel_plot</c>, by passing it
the fxt trace:

\verbatim
$ starpu_perfmodel_plot -s non_linear_memset_regression_based -i /tmp/prof_file_foo_0
\endverbatim

It will produce a <c>.gp</c> file which contains both the performance model
curves, and the profiling measurements.

\image html starpu_non_linear_memset_regression_based_2.png
\image latex starpu_non_linear_memset_regression_based_2.eps "" width=\textwidth

If you have the statistical tool <c>R</c> installed, you can additionally use

\verbatim
$ starpu_codelet_histo_profile distrib.data
\endverbatim

Which will create one <c>.pdf</c> file per codelet and per input size, showing a
histogram of the codelet execution time distribution.

\image html distrib_data_histo.png
\image latex distrib_data_histo.eps "" width=\textwidth

\section TraceStatistics Trace Statistics

More than just codelet performance, it is interesting to get statistics over all
kinds of StarPU states (allocations, data transfers, etc.). This is particularly
useful to check what may have gone wrong in the accurracy of the simgrid
simulation.

This requires the <c>R</c> statistical tool, with the <c>plyr</c>,
<c>ggplot2</c> and <c>data.table</c> packages. If your system
distribution does not have packages for these, one can fetch them from
<c>CRAN</c>:

\verbatim
$ R
> install.packages("plyr")
> install.packages("ggplot2")
> install.packages("data.table")
> install.packages("knitr")
\endverbatim

The <c>pj_dump</c> tool from <c>pajeng</c> is also needed (see
https://github.com/schnorr/pajeng)

One can then get textual or <c>.csv</c> statistics over the trace states:

\verbatim
$ starpu_paje_state_stats -v native.trace simgrid.trace
"Value"         "Events_native.csv" "Duration_native.csv" "Events_simgrid.csv" "Duration_simgrid.csv"
"Callback"      220                 0.075978              220                  0
"chol_model_11" 10                  565.176               10                   572.8695
"chol_model_21" 45                  9184.828              45                   9170.719
"chol_model_22" 165                 64712.07              165                  64299.203
$ starpu_paje_state_stats native.trace simgrid.trace
\endverbatim

An other way to get statistics of StarPU states (without installing <c>R</c> and
<c>pj_dump</c>) is to use the <c>starpu_trace_state_stats.py</c> script which parses the
generated <c>trace.rec</c> file instead of the <c>paje.trace</c> file. The output is similar
to the previous script but it doesn't need any dependencies.

The different prefixes used in <c>trace.rec</c> are:

\verbatim
E: Event type
N: Event name
C: Event category
W: Worker ID
T: Thread ID
S: Start time
\endverbatim

Here's an example on how to use it:

\verbatim
$ python starpu_trace_state_stats.py trace.rec | column -t -s ","
"Name"		"Count" "Type"	"Duration"
"Callback"       220	Runtime	0.075978
"chol_model_11"  10	Task	565.176
"chol_model_21"  45	Task	9184.828
"chol_model_22"  165	Task	64712.07
\endverbatim

<c>starpu_trace_state_stats.py</c> can also be used to compute the different
efficiencies. Refer to the usage description to show some examples.

And one can plot histograms of execution times, of several states for instance:
\verbatim
$ starpu_paje_draw_histogram -n chol_model_11,chol_model_21,chol_model_22 native.trace simgrid.trace
\endverbatim

and see the resulting pdf file:

\image html paje_draw_histogram.png
\image latex paje_draw_histogram.eps "" width=\textwidth

A quick statistical report can be generated by using:

\verbatim
$ starpu_paje_summary native.trace simgrid.trace
\endverbatim

it includes gantt charts, execution summaries, as well as state duration charts
and time distribution histograms.

Other external Paje analysis tools can be used on these traces, one just needs
to sort the traces by timestamp order (which not guaranteed to make recording
more efficient):

\verbatim
$ starpu_paje_sort paje.trace
\endverbatim

\section TheoreticalLowerBoundOnExecutionTime Theoretical Lower Bound On Execution Time

StarPU can record a trace of what tasks are needed to complete the
application, and then, by using a linear system, provide a theoretical lower
bound of the execution time (i.e. with an ideal scheduling).

The computed bound is not really correct when not taking into account
dependencies, but for an application which have enough parallelism, it is very
near to the bound computed with dependencies enabled (which takes a huge lot
more time to compute), and thus provides a good-enough estimation of the ideal
execution time.

\ref TheoreticalLowerBoundOnExecutionTimeExample provides an example on how to
use this.

\section TheoreticalLowerBoundOnExecutionTimeExample Theoretical Lower Bound On Execution Time Example

For kernels with history-based performance models (and provided that
they are completely calibrated), StarPU can very easily provide a
theoretical lower bound for the execution time of a whole set of
tasks. See for instance <c>examples/lu/lu_example.c</c>: before
submitting tasks, call the function starpu_bound_start(), and after
complete execution, call starpu_bound_stop().
starpu_bound_print_lp() or starpu_bound_print_mps() can then be used
to output a Linear Programming problem corresponding to the schedule
of your tasks. Run it through <c>lp_solve</c> or any other linear
programming solver, and that will give you a lower bound for the total
execution time of your tasks. If StarPU was compiled with the library
<c>glpk</c> installed, starpu_bound_compute() can be used to solve it
immediately and get the optimized minimum, in ms. Its parameter
<c>integer</c> allows to decide whether integer resolution should be
computed and returned

The <c>deps</c> parameter tells StarPU whether to take tasks, implicit
data, and tag dependencies into account. Tags released in a callback
or similar are not taken into account, only tags associated with a task are.
It must be understood that the linear programming
problem size is quadratic with the number of tasks and thus the time to solve it
will be very long, it could be minutes for just a few dozen tasks. You should
probably use <c>lp_solve -timeout 1 test.pl -wmps test.mps</c> to convert the
problem to MPS format and then use a better solver, <c>glpsol</c> might be
better than <c>lp_solve</c> for instance (the <c>--pcost</c> option may be
useful), but sometimes doesn't manage to converge. <c>cbc</c> might look
slower, but it is parallel. For <c>lp_solve</c>, be sure to try at least all the
<c>-B</c> options. For instance, we often just use <c>lp_solve -cc -B1 -Bb
-Bg -Bp -Bf -Br -BG -Bd -Bs -BB -Bo -Bc -Bi</c> , and the <c>-gr</c> option can
also be quite useful. The resulting schedule can be observed by using
the tool <c>starpu_lp2paje</c>, which converts it into the Paje
format.

Data transfer time can only be taken into account when <c>deps</c> is set. Only
data transfers inferred from implicit data dependencies between tasks are taken
into account. Other data transfers are assumed to be completely overlapped.

Setting <c>deps</c> to 0 will only take into account the actual computations
on processing units. It however still properly takes into account the varying
performances of kernels and processing units, which is quite more accurate than
just comparing StarPU performances with the fastest of the kernels being used.

The <c>prio</c> parameter tells StarPU whether to simulate taking into account
the priorities as the StarPU scheduler would, i.e. schedule prioritized
tasks before less prioritized tasks, to check to which extend this results
to a less optimal solution. This increases even more computation time.

\section MemoryFeedback Memory Feedback

It is possible to enable memory statistics. To do so, you need to pass
the option \ref enable-memory-stats "--enable-memory-stats" when running <c>configure</c>. It is then
possible to call the function starpu_data_display_memory_stats() to
display statistics about the current data handles registered within StarPU.

Moreover, statistics will be displayed at the end of the execution on
data handles which have not been cleared out. This can be disabled by
setting the environment variable \ref STARPU_MEMORY_STATS to <c>0</c>.

For example, if you do not unregister data at the end of the complex
example, you will get something similar to:

\verbatim
$ STARPU_MEMORY_STATS=0 ./examples/interface/complex
Complex[0] = 45.00 + 12.00 i
Complex[0] = 78.00 + 78.00 i
Complex[0] = 45.00 + 12.00 i
Complex[0] = 45.00 + 12.00 i
\endverbatim

\verbatim
$ STARPU_MEMORY_STATS=1 ./examples/interface/complex
Complex[0] = 45.00 + 12.00 i
Complex[0] = 78.00 + 78.00 i
Complex[0] = 45.00 + 12.00 i
Complex[0] = 45.00 + 12.00 i

#---------------------
Memory stats:
#-------
Data on Node #3
#-----
Data : 0x553ff40
Size : 16

#--
Data access stats
/!\ Work Underway
Node #0
	Direct access : 4
	Loaded (Owner) : 0
	Loaded (Shared) : 0
	Invalidated (was Owner) : 0

Node #3
	Direct access : 0
	Loaded (Owner) : 0
	Loaded (Shared) : 1
	Invalidated (was Owner) : 0

#-----
Data : 0x5544710
Size : 16

#--
Data access stats
/!\ Work Underway
Node #0
	Direct access : 2
	Loaded (Owner) : 0
	Loaded (Shared) : 1
	Invalidated (was Owner) : 1

Node #3
	Direct access : 0
	Loaded (Owner) : 1
	Loaded (Shared) : 0
	Invalidated (was Owner) : 0
\endverbatim

\section DataStatistics Data Statistics

Different data statistics can be displayed at the end of the execution
of the application. To enable them, you need to define the environment
variable \ref STARPU_ENABLE_STATS. When calling
starpu_shutdown() various statistics will be displayed,
execution, MSI cache statistics, allocation cache statistics, and data
transfer statistics. The display can be disabled by setting the
environment variable \ref STARPU_STATS to <c>0</c>.

\verbatim
$ ./examples/cholesky/cholesky_tag
Computation took (in ms)
518.16
Synthetic GFlops : 44.21
#---------------------
MSI cache stats :
TOTAL MSI stats	hit 1622 (66.23 %)	miss 827 (33.77 %)
...
\endverbatim

\verbatim
$ STARPU_STATS=0 ./examples/cholesky/cholesky_tag
Computation took (in ms)
518.16
Synthetic GFlops : 44.21
\endverbatim

// TODO: data transfer stats are similar to the ones displayed when
// setting STARPU_BUS_STATS

*/