starpu-max/doc/doxygen/chapters/370_online_performance_tools.doxy

/* StarPU --- Runtime system for heterogeneous multicore architectures.
 *
 * Copyright (C) 2010-2017                                CNRS
 * Copyright (C) 2011-2012,2016                           Inria
 * Copyright (C) 2009-2011,2014,2016                      Université de Bordeaux
 *
 * StarPU is free software; you can redistribute it and/or modify
 * it under the terms of the GNU Lesser General Public License as published by
 * the Free Software Foundation; either version 2.1 of the License, or (at
 * your option) any later version.
 *
 * StarPU is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
 *
 * See the GNU Lesser General Public License in COPYING.LGPL for more details.
 */

/*! \page OnlinePerformanceTools Online Performance Tools

\section On-linePerformanceFeedback On-line Performance Feedback

\subsection EnablingOn-linePerformanceMonitoring Enabling On-line Performance Monitoring

In order to enable online performance monitoring, the application can
call starpu_profiling_status_set() with the parameter
::STARPU_PROFILING_ENABLE. It is possible to detect whether monitoring
is already enabled or not by calling starpu_profiling_status_get().
Enabling monitoring also reinitialize all previously collected
feedback. The environment variable \ref STARPU_PROFILING can also be
set to <c>1</c> to achieve the same effect. The function
starpu_profiling_init() can also be called during the execution to
reinitialize performance counters and to start the profiling if the
environment variable \ref STARPU_PROFILING is set to <c>1</c>.

Likewise, performance monitoring is stopped by calling
starpu_profiling_status_set() with the parameter
::STARPU_PROFILING_DISABLE. Note that this does not reset the
performance counters so that the application may consult them later
on.

More details about the performance monitoring API are available in \ref API_Profiling.

\subsection Per-taskFeedback Per-task Feedback

If profiling is enabled, a pointer to a structure
starpu_profiling_task_info is put in the field
starpu_task::profiling_info when a task terminates. This structure is
automatically destroyed when the task structure is destroyed, either
automatically or by calling starpu_task_destroy().

The structure starpu_profiling_task_info indicates the date when the
task was submitted (starpu_profiling_task_info::submit_time), started
(starpu_profiling_task_info::start_time), and terminated
(starpu_profiling_task_info::end_time), relative to the initialization
of StarPU with starpu_init(). It also specifies the identifier of the worker
that has executed the task (starpu_profiling_task_info::workerid).
These date are stored as <c>timespec</c> structures which the user may convert
into micro-seconds using the helper function
starpu_timing_timespec_to_us().

It it worth noting that the application may directly access this structure from
the callback executed at the end of the task. The structure starpu_task
associated to the callback currently being executed is indeed accessible with
the function starpu_task_get_current().

\subsection Per-codeletFeedback Per-codelet Feedback

The field starpu_codelet::per_worker_stats is
an array of counters. The <c>i</c>-th entry of the array is incremented every time a
task implementing the codelet is executed on the <c>i</c>-th worker.
This array is not reinitialized when profiling is enabled or disabled.

\subsection Per-workerFeedback Per-worker Feedback

The second argument returned by the function
starpu_profiling_worker_get_info() is a structure
starpu_profiling_worker_info that gives statistics about the specified
worker. This structure specifies when StarPU started collecting
profiling information for that worker
(starpu_profiling_worker_info::start_time), the
duration of the profiling measurement interval
(starpu_profiling_worker_info::total_time), the time spent executing
kernels (starpu_profiling_worker_info::executing_time), the time
spent sleeping because there is no task to execute at all
(starpu_profiling_worker_info::sleeping_time), and the number of tasks that were executed
while profiling was enabled. These values give an estimation of the
proportion of time spent do real work, and the time spent either
sleeping because there are not enough executable tasks or simply
wasted in pure StarPU overhead.

Calling starpu_profiling_worker_get_info() resets the profiling
information associated to a worker.

To easily display all this information, the environment variable
\ref STARPU_WORKER_STATS can be set to <c>1</c> (in addition to setting
\ref STARPU_PROFILING to 1). A summary will then be displayed at program termination:

\verbatim
Worker stats:
CUDA 0.0 (4.7 GiB)
	480 task(s)
	total: 1574.82 ms executing: 1510.72 ms sleeping: 0.00 ms overhead 64.10 ms
	325.217970 GFlop/s

CPU 0
	22 task(s)
	total: 1574.82 ms executing: 1364.81 ms sleeping: 0.00 ms overhead 210.01 ms
	7.512057 GFlop/s

CPU 1
	14 task(s)
	total: 1574.82 ms executing: 1500.13 ms sleeping: 0.00 ms overhead 74.69 ms
	6.675853 GFlop/s

CPU 2
	14 task(s)
	total: 1574.82 ms executing: 1553.12 ms sleeping: 0.00 ms overhead 21.70 ms
	7.152886 GFlop/s
\endverbatim

The number of GFlops is available because the starpu_task::flops field of the
tasks were filled (or \ref STARPU_FLOPS used in starpu_task_insert()).

When an FxT trace is generated (see \ref GeneratingTracesWithFxT), it is also
possible to use the tool <c>starpu_workers_activity</c> (see
\ref MonitoringActivity) to generate a graphic showing the evolution of
these values during the time, for the different workers.

\subsection Bus-relatedFeedback Bus-related Feedback

// how to enable/disable performance monitoring
// what kind of information do we get ?

The bus speed measured by StarPU can be displayed by using the tool
<c>starpu_machine_display</c>, for instance:

\verbatim
StarPU has found:
        3 CUDA devices
                CUDA 0 (Tesla C2050 02:00.0)
                CUDA 1 (Tesla C2050 03:00.0)
                CUDA 2 (Tesla C2050 84:00.0)
from    to RAM          to CUDA 0       to CUDA 1       to CUDA 2
RAM     0.000000        5176.530428     5176.492994     5191.710722
CUDA 0  4523.732446     0.000000        2414.074751     2417.379201
CUDA 1  4523.718152     2414.078822     0.000000        2417.375119
CUDA 2  4534.229519     2417.069025     2417.060863     0.000000
\endverbatim

Statistics about the data transfers which were performed and temporal average
of bandwidth usage can be obtained by setting the environment variable
\ref STARPU_BUS_STATS to <c>1</c>; a summary will then be displayed at program termination:

\verbatim
Data transfer stats:
	RAM 0 -> CUDA 0	319.92 MB	213.10 MB/s	(transfers : 91 - avg 3.52 MB)
	CUDA 0 -> RAM 0	214.45 MB	142.85 MB/s	(transfers : 61 - avg 3.52 MB)
	RAM 0 -> CUDA 1	302.34 MB	201.39 MB/s	(transfers : 86 - avg 3.52 MB)
	CUDA 1 -> RAM 0	133.59 MB	88.99 MB/s	(transfers : 38 - avg 3.52 MB)
	CUDA 0 -> CUDA 1	144.14 MB	96.01 MB/s	(transfers : 41 - avg 3.52 MB)
	CUDA 1 -> CUDA 0	130.08 MB	86.64 MB/s	(transfers : 37 - avg 3.52 MB)
	RAM 0 -> CUDA 2	312.89 MB	208.42 MB/s	(transfers : 89 - avg 3.52 MB)
	CUDA 2 -> RAM 0	133.59 MB	88.99 MB/s	(transfers : 38 - avg 3.52 MB)
	CUDA 0 -> CUDA 2	151.17 MB	100.69 MB/s	(transfers : 43 - avg 3.52 MB)
	CUDA 2 -> CUDA 0	105.47 MB	70.25 MB/s	(transfers : 30 - avg 3.52 MB)
	CUDA 1 -> CUDA 2	175.78 MB	117.09 MB/s	(transfers : 50 - avg 3.52 MB)
	CUDA 2 -> CUDA 1	203.91 MB	135.82 MB/s	(transfers : 58 - avg 3.52 MB)
Total transfers: 2.27 GB
\endverbatim

\subsection StarPU-TopInterface StarPU-Top Interface

StarPU-Top is an interface which remotely displays the on-line state of a StarPU
application and permits the user to change parameters on the fly.

Variables to be monitored can be registered by calling the functions
starpu_top_add_data_boolean(), starpu_top_add_data_integer(),
starpu_top_add_data_float(), e.g.:

\code{.c}
starpu_top_data *data = starpu_top_add_data_integer("mynum", 0, 100, 1);
\endcode

The application should then call starpu_top_init_and_wait() to give its name
and wait for StarPU-Top to get a start request from the user. The name is used
by StarPU-Top to quickly reload a previously-saved layout of parameter display.

\code{.c}
starpu_top_init_and_wait("the application");
\endcode

The new values can then be provided thanks to
starpu_top_update_data_boolean(), starpu_top_update_data_integer(),
starpu_top_update_data_float(), e.g.:

\code{.c}
starpu_top_update_data_integer(data, mynum);
\endcode

Updateable parameters can be registered thanks to starpu_top_register_parameter_boolean(), starpu_top_register_parameter_integer(), starpu_top_register_parameter_float(), e.g.:

\code{.c}
float alpha;
starpu_top_register_parameter_float("alpha", &alpha, 0, 10, modif_hook);
\endcode

<c>modif_hook</c> is a function which will be called when the parameter is being modified, it can for instance print the new value:

\code{.c}
void modif_hook(struct starpu_top_param *d)
{
    fprintf(stderr,"%s has been modified: %f\n", d->name, alpha);
}
\endcode

Task schedulers should notify StarPU-Top when it has decided when a task will be
scheduled, so that it can show it in its Gantt chart, for instance:

\code{.c}
starpu_top_task_prevision(task, workerid, begin, end);
\endcode

Starting StarPU-Top (StarPU-Top is started via the binary
<c>starpu_top</c>) and the application can be done in two ways:

<ul>
<li> The application is started by hand on some machine (and thus already
waiting for the start event). In the Preference dialog of StarPU-Top, the SSH
checkbox should be unchecked, and the hostname and port (default is 2011) on
which the application is already running should be specified. Clicking on the
connection button will thus connect to the already-running application.
</li>
<li> StarPU-Top is started first, and clicking on the connection button will
start the application itself (possibly on a remote machine). The SSH checkbox
should be checked, and a command line provided, e.g.:

\verbatim
$ ssh myserver STARPU_SCHED=dmda ./application
\endverbatim

If port 2011 of the remote machine can not be accessed directly, an ssh port bridge should be added:

\verbatim
$ ssh -L 2011:localhost:2011 myserver STARPU_SCHED=dmda ./application
\endverbatim

and "localhost" should be used as IP Address to connect to.
</li>
</ul>

\section TaskAndWorkerProfiling Task And Worker Profiling

A full example showing how to use the profiling API is available in
the StarPU sources in the directory <c>examples/profiling/</c>.

\code{.c}
struct starpu_task *task = starpu_task_create();
task->cl = &cl;
task->synchronous = 1;
/* We will destroy the task structure by hand so that we can
 * query the profiling info before the task is destroyed. */
task->destroy = 0;

/* Submit and wait for completion (since synchronous was set to 1) */
starpu_task_submit(task);

/* The task is finished, get profiling information */
struct starpu_profiling_task_info *info = task->profiling_info;

/* How much time did it take before the task started ? */
double delay += starpu_timing_timespec_delay_us(&info->submit_time, &info->start_time);

/* How long was the task execution ? */
double length += starpu_timing_timespec_delay_us(&info->start_time, &info->end_time);

/* We no longer need the task structure */
starpu_task_destroy(task);
\endcode

\code{.c}
/* Display the occupancy of all workers during the test */
int worker;
for (worker = 0; worker < starpu_worker_get_count(); worker++)
{
        struct starpu_profiling_worker_info worker_info;
        int ret = starpu_profiling_worker_get_info(worker, &worker_info);
        STARPU_ASSERT(!ret);

        double total_time = starpu_timing_timespec_to_us(&worker_info.total_time);
        double executing_time = starpu_timing_timespec_to_us(&worker_info.executing_time);
        double sleeping_time = starpu_timing_timespec_to_us(&worker_info.sleeping_time);
        double overhead_time = total_time - executing_time - sleeping_time;

        float executing_ratio = 100.0*executing_time/total_time;
        float sleeping_ratio = 100.0*sleeping_time/total_time;
        float overhead_ratio = 100.0 - executing_ratio - sleeping_ratio;

        char workername[128];
        starpu_worker_get_name(worker, workername, 128);
        fprintf(stderr, "Worker %s:\n", workername);
        fprintf(stderr, "\ttotal time: %.2lf ms\n", total_time*1e-3);
        fprintf(stderr, "\texec time: %.2lf ms (%.2f %%)\n",
                executing_time*1e-3, executing_ratio);
        fprintf(stderr, "\tblocked time: %.2lf ms (%.2f %%)\n",
                sleeping_time*1e-3, sleeping_ratio);
        fprintf(stderr, "\toverhead time: %.2lf ms (%.2f %%)\n",
                overhead_time*1e-3, overhead_ratio);
}
\endcode

\section PerformanceModelExample Performance Model Example

To achieve good scheduling, StarPU scheduling policies need to be able to
estimate in advance the duration of a task. This is done by giving to codelets
a performance model, by defining a structure starpu_perfmodel and
providing its address in the field starpu_codelet::model. The fields
starpu_perfmodel::symbol and starpu_perfmodel::type are mandatory, to
give a name to the model, and the type of the model, since there are
several kinds of performance models. For compatibility, make sure to
initialize the whole structure to zero, either by using explicit
memset(), or by letting the compiler implicitly do it as examplified
below.

<ul>
<li>
Measured at runtime (model type ::STARPU_HISTORY_BASED). This assumes that for a
given set of data input/output sizes, the performance will always be about the
same. This is very true for regular kernels on GPUs for instance (<0.1% error),
and just a bit less true on CPUs (~=1% error). This also assumes that there are
few different sets of data input/output sizes. StarPU will then keep record of
the average time of previous executions on the various processing units, and use
it as an estimation. History is done per task size, by using a hash of the input
and ouput sizes as an index.
It will also save it in <c>$STARPU_HOME/.starpu/sampling/codelets</c>
for further executions, and can be observed by using the tool
<c>starpu_perfmodel_display</c>, or drawn by using
the tool <c>starpu_perfmodel_plot</c> (\ref PerformanceModelCalibration).  The
models are indexed by machine name. To
share the models between machines (e.g. for a homogeneous cluster), use
<c>export STARPU_HOSTNAME=some_global_name</c>. Measurements are only done
when using a task scheduler which makes use of it, such as
<c>dmda</c>. Measurements can also be provided explicitly by the application, by
using the function starpu_perfmodel_update_history().

The following is a small code example.

If e.g. the code is recompiled with other compilation options, or several
variants of the code are used, the <c>symbol</c> string should be changed to reflect
that, in order to recalibrate a new model from zero. The <c>symbol</c> string can even
be constructed dynamically at execution time, as long as this is done before
submitting any task using it.

\code{.c}
static struct starpu_perfmodel mult_perf_model =
{
    .type = STARPU_HISTORY_BASED,
    .symbol = "mult_perf_model"
};

struct starpu_codelet cl =
{
    .cpu_funcs = { cpu_mult },
    .cpu_funcs_name = { "cpu_mult" },
    .nbuffers = 3,
    .modes = { STARPU_R, STARPU_R, STARPU_W },
    /* for the scheduling policy to be able to use performance models */
    .model = &mult_perf_model
};
\endcode

</li>
<li>
Measured at runtime and refined by regression (model types
::STARPU_REGRESSION_BASED and ::STARPU_NL_REGRESSION_BASED). This
still assumes performance regularity, but works
with various data input sizes, by applying regression over observed
execution times. ::STARPU_REGRESSION_BASED uses an <c>a*n^b</c> regression
form, ::STARPU_NL_REGRESSION_BASED uses an <c>a*n^b+c</c> (more precise than
::STARPU_REGRESSION_BASED, but costs a lot more to compute).

For instance,
<c>tests/perfmodels/regression_based.c</c> uses a regression-based performance
model for the function memset().

Of course, the application has to issue
tasks with varying size so that the regression can be computed. StarPU will not
trust the regression unless there is at least 10% difference between the minimum
and maximum observed input size. It can be useful to set the
environment variable \ref STARPU_CALIBRATE to <c>1</c> and run the application
on varying input sizes with \ref STARPU_SCHED set to <c>dmda</c> scheduler,
so as to feed the performance model for a variety of
inputs. The application can also provide the measurements explictly by
using the function starpu_perfmodel_update_history(). The tools
<c>starpu_perfmodel_display</c> and <c>starpu_perfmodel_plot</c> can
be used to observe how much the performance model is calibrated
(\ref PerformanceModelCalibration); when their output look good,
\ref STARPU_CALIBRATE can be reset to <c>0</c> to let
StarPU use the resulting performance model without recording new measures, and
\ref STARPU_SCHED can be set to <c>dmda</c> to benefit from the performance models. If
the data input sizes vary a lot, it is really important to set
\ref STARPU_CALIBRATE to <c>0</c>, otherwise StarPU will continue adding the
measures, and result with a very big performance model, which will take time a
lot of time to load and save.

For non-linear regression, since computing it
is quite expensive, it is only done at termination of the application. This
means that the first execution of the application will use only history-based
performance model to perform scheduling, without using regression.
</li>

<li>

Another type of model is ::STARPU_MULTIPLE_REGRESSION_BASED, which
is based on multiple linear regression. In this model, the user
defines both the relevant parameters and the equation for computing the
task duration.


\f[
T_{kernel} = a + b(M^{\alpha_1} * N^{\beta_1} * K^{\gamma_1}) + c(M^{\alpha_2} * N^{\beta_2} * K^{\gamma_2}) + ...
\f]


\f$M, N, K\f$ are the parameters of the task, added at the task
creation. These need to be extracted by the <c>cl_perf_func</c>
function, which should be defined by the user. \f$\alpha, \beta,
\gamma\f$ are the exponents defined by the user in
<c>model->combinations</c> table. Finally, coefficients \f$a, b, c\f$
are computed automatically by the StarPU at the end of the execution, using least
squares method of the <c>dgels_</c> LAPACK function.

<c>examples/mlr/mlr.c</c> example provides more details on
the usage of ::STARPU_MULTIPLE_REGRESSION_BASED models.

Coefficients computation is done at the end of the execution, and the
results are stored in standard codelet perfmodel files. Additional
files containing the duration of task together with the value of each
parameter are stored in <c>.starpu/sampling/codelets/tmp/</c>
directory. These files are reused when \ref STARPU_CALIBRATE
environment variable is set to <c>1</c>, to recompute coefficients
based on the current, but also on the previous
executions. Additionally, when multiple linear regression models are
disabled (using "--disable-mlr" configuration option) or when the
<c>model->combinations</c> are not defined, StarPU will still write
output files into <c>.starpu/sampling/codelets/tmp/</c> to allow
performing an analysis. This analysis typically aims at finding the
most appropriate equation for the codelet and
<c>tools/starpu_mlr_analysis</c> script provides an example of how to
perform such study.

</li>

<li>
Provided as an estimation from the application itself (model type
::STARPU_COMMON and field starpu_perfmodel::cost_function),
see for instance
<c>examples/common/blas_model.h</c> and <c>examples/common/blas_model.c</c>.
</li>

<li>
Provided explicitly by the application (model type ::STARPU_PER_ARCH):
either field starpu_perfmodel::arch_cost_function, or
the fields <c>.per_arch[arch][nimpl].cost_function</c> have to be
filled with pointers to functions which return the expected duration
of the task in micro-seconds, one per architecture, see for instance
<c>tests/datawizard/locality.c</c>
</li>
</ul>

For ::STARPU_HISTORY_BASED, ::STARPU_REGRESSION_BASED, and
::STARPU_NL_REGRESSION_BASED, the dimensions of task data (both input
and output) are used as an index by default. ::STARPU_HISTORY_BASED uses a CRC
hash of the dimensions as an index to distinguish histories, and
::STARPU_REGRESSION_BASED and ::STARPU_NL_REGRESSION_BASED use the total
size as an index for the regression.

The starpu_perfmodel::size_base and starpu_perfmodel::footprint fields however
permit the application to override that, when for instance some of the data
do not matter for task cost (e.g. mere reference table), or when using sparse
structures (in which case it is the number of non-zeros which matter), or when
there is some hidden parameter such as the number of iterations, or when the
application actually has a very good idea of the complexity of the algorithm,
and just not the speed of the processor, etc.  The example in the directory
<c>examples/pi</c> uses this to include the number of iterations in the base
size. starpu_perfmodel::size_base should be used when the variance of the actual
performance is known (i.e. bigger returned value is longer execution
time), and thus particularly useful for ::STARPU_REGRESSION_BASED or
::STARPU_NL_REGRESSION_BASED. starpu_perfmodel::footprint can be used when the
variance of the actual performance is unknown (irregular performance behavior,
etc.), and thus only useful for ::STARPU_HISTORY_BASED.
starpu_task_data_footprint() can be used as a base and combined with other
parameters through starpu_hash_crc32c_be() for instance.

StarPU will automatically determine when the performance model is calibrated,
or rather, it will assume the performance model is calibrated until the
application submits a task for which the performance can not be predicted. For
::STARPU_HISTORY_BASED, StarPU will require 10 (STARPU_CALIBRATE_MINIMUM)
measurements for a given size before estimating that an average can be taken as
estimation for further executions with the same size. For
::STARPU_REGRESSION_BASED and ::STARPU_NL_REGRESSION_BASED, StarPU will require
10 (STARPU_CALIBRATE_MINIMUM) measurements, and that the minimum measured
data size is smaller than 90% of the maximum measured data size (i.e. the
measurement interval is large enough for a regression to have a meaning).
Calibration can also be forced by setting the \ref STARPU_CALIBRATE environment
variable to <c>1</c>, or even reset by setting it to <c>2</c>.

How to use schedulers which can benefit from such performance model is explained
in \ref TaskSchedulingPolicy.

The same can be done for task energy consumption estimation, by setting
the field starpu_codelet::energy_model the same way as the field
starpu_codelet::model. Note: for now, the application has to give to
the energy consumption performance model a name which is different from
the execution time performance model.

The application can request time estimations from the StarPU performance
models by filling a task structure as usual without actually submitting
it. The data handles can be created by calling any of the functions
<c>starpu_*_data_register</c> with a <c>NULL</c> pointer and <c>-1</c>
node and the desired data sizes, and need to be unregistered as usual.
The functions starpu_task_expected_length() and
starpu_task_expected_energy() can then be called to get an estimation
of the task cost on a given arch. starpu_task_footprint() can also be
used to get the footprint used for indexing history-based performance
models. starpu_task_destroy() needs to be called to destroy the dummy
task afterwards. See <c>tests/perfmodels/regression_based.c</c> for an example.

\section DataTrace Data trace and tasks length
It is possible to get statistics about tasks length and data size by using :
\verbatim
$ starpu_fxt_data_trace filename [codelet1 codelet2 ... codeletn]
\endverbatim
Where filename is the FxT trace file and codeletX the names of the codelets you
want to profile (if no names are specified, <c>starpu_fxt_data_trace</c> will profile them all).
This will create a file, <c>data_trace.gp</c> which
can be executed to get a <c>.eps</c> image of these results. On the image, each point represents a
task, and each color corresponds to a codelet.

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

*/