lines 5-129 of file: xrst/csv_column.xrst

{xrst_begin csv_column}
{xrst_spell
   dd
   kutta
   mm
   newline
   runge
   ture
   yyyy
}

The cmpad Csv File Columns
##########################

#. The comma character is used to terminate each column.
#. The newline character is used to terminate each row.
#. The first line (the header)
   contains the column names in the following order::
   {xrst_comment BEGIN HEADER_LINE}
   rate,min_time,package,algorithm,n_arg,date,compiler,debug,language
   {xrst_comment END HEADER_LINE}


rate
****
This floating point value is the number of times per second that the algorithm,
or a derivative of the algorithm, was calculated.

min_time
********
This floating point value is the minimum time
in seconds to average the computation rate over.
The computation is repeated enough times so that this minimum time is reached.

package
*******
This string is the name of the AD package that *rate* corresponds to.
It can be any of the following values:
{xrst_literal ,
   bin/get_package.sh
   # BEGIN CPP_PACKAGE_LIST, # END CPP_PACKAGE_LIST
   # BEGIN PYTHON_PACKAGE_LIST, # END PYTHON_PACKAGE_LIST
}

none
====
If *package* is ``none`` , this is not a derivative calculation but rather
an evaluations of the function using the algorithm with type ``double`` .

algorithm
*********
The algorithm used by this test.
The possible values (so far) are
:ref:`det_by_minor-name` ,
:ref:`an_ode-name` , and
:ref:`llsq_obj-name` .

n_arg
*****
This is the value of then *n_arg* integer and is always the dimension of
the argument space for the function and algorithm; see the heading n_arg in
:ref:`det_by_minor <det_by_minor@option@n_arg>` ,
:ref:`an_ode <an_ode@option@n_arg>` ,
:ref:`llsq_obj <llsq_obj@option@n_arg>`  .



n_other
*******
the meaning of this integer is algorithm specific:

.. csv-table::
   :header-rows: 1

   *algorithm*, *n_other*
   det_by_minor, must be zero and is not used
   an_ode, number of Runge-Kutta steps used to approximate ODE
   llsq_obj, is the number of data points in the least squares fit.

For more information see the heading n_other in
:ref:`det_by_minor <det_by_minor@option@n_other>` ,
:ref:`an_ode <an_ode@option@n_other>` ,
:ref:`llsq_obj <llsq_obj@option@n_other>`  .

time_setup
**********
If *time_setup* is ``true`` ( ``false`` ) ,
the setup time was included (was not included)
during the speed testing for this algorithm.
Some AD packages may spend more setup time to use less evaluation time
for each new argument value.

date
****
This is the date when the timing test was run in yyyy-mm-dd format.

compiler
********
This is a string identifying the version of the C++ or python
used to compile or interpret the algorithm.
2DO: We should ensure that same compiler is used to compile
libraries used by the AD package.

debug
*****
If this is ``ture`` ( ``false`` ),
the algorithm, and AD package libraries, where compiled for debugging
(for release).
If it is the empty string, there is no choice for compiling the algorithm
or AD package; e.g., if *package* is ``none`` and *language* is ``python`` ,
then *debug* is empty.

language
********
This is the language used to compile or interpret the algorithm and
its derivatives. The possible values so far are ``python`` and ``c++`` .

special
*******
If this is ``true`` ,
a special version of the algorithm for this particular *package* was used.
Otherwise it is ``false`` and the algorithm can be used
by any of the packages for this *language* .

{xrst_end csv_column}