/* $Id: multi_thread.omh 2765 2013-03-03 15:48:35Z bradbell $ */
/* --------------------------------------------------------------------------
CppAD: C++ Algorithmic Differentiation: Copyright (C) 2003-13 Bradley M. Bell

CppAD is distributed under multiple licenses. This distribution is under
the terms of the 
                    Eclipse Public License Version 1.0.

A copy of this license is included in the COPYING file of this distribution.
Please visit http://www.coin-or.org/CppAD/ for information on other licenses.
-------------------------------------------------------------------------- */
$begin multi_thread$$
$spell
	cppad_max_num_threads
	alloc
	CppAD
	Rosen
	Runge
$$

$section Using CppAD in a Multi-Threading Environment$$
$index parallel, mode$$
$index execution, parallel$$
$index mode, parallel$$
$index multi-threading$$

$head Purpose$$
Extra steps and care must be taken to use CppAD 
in $cref/parallel/ta_in_parallel/$$ execution mode. 
This section collects this information in one place.

$head CPPAD_MAX_NUM_THREADS$$
$index maximum, number of threads$$
$index threads, maximum number of$$
$index CPPAD_MAX_NUM_THREADS$$
The value $code CPPAD_MAX_NUM_THREADS$$ is an absolute maximum for 
the number of threads that CppAD should support.
If this preprocessor symbol is defined
before including any CppAD header files,
it must be an integer greater than or equal to one.
Otherwise, $cref/cppad_max_num_threads/cmake/cppad_max_num_threads/$$
is used to define this preprocessor symbol.
Note that the minimum allowable value for
$icode cppad_max_num_threads$$ is 4; i.e., you can only
get smaller values for $code CPPAD_MAX_NUM_THREADS$$ by
defining it before including the CppAD header files.

$head parallel_setup$$
Using any of the following routines in a multi-threading environment
requires that $cref/thread_alloc::parallel_setup/ta_parallel_setup/$$ 
has been completed:
$cref/CppAD::vector/CppAD_vector/Memory and Parallel Mode/$$,
$cref/CheckSimpleVector/CheckSimpleVector/Parallel Mode/$$,
$cref/CheckNumericType/CheckNumericType/Parallel Mode/$$,
$cref parallel_ad$$.

$head hold_memory$$
Memory allocation should be much faster after calling
$code hold_memory$$ with
$cref/value/ta_hold_memory/value/$$ equal to true. 
This may even be true if there is only one thread.

$head Parallel AD$$
One must first call
$cref/thread_alloc::parallel_setup/ta_parallel_setup/$$ 
and then call $cref parallel_ad$$ before using $code AD$$ types
in $cref/parallel/ta_in_parallel/$$ execution mode.

$head Initialization$$
The following routines must be called at least once before being used
in parallel mode:
$cref/CheckSimpleVector/CheckSimpleVector/Parallel Mode/$$,
$cref/CheckNumericType/CheckNumericType/Parallel Mode/$$,
$cref/discrete functions/Discrete/Parallel Mode/$$,
$cref/Rosen34/Rosen34/Parallel Mode/$$,
$cref/Runge45/Runge45/Parallel Mode/$$.

$head Same Thread$$
Some operations must be preformed by the same thread:
$cref/ADFun/FunConstruct/Parallel Mode/$$,
$cref Independent$$, $cref Dependent$$.


$head Parallel Prohibited$$
The following routine cannot be called in parallel mode:
$cref/ErrorHandler constructor/ErrorHandler/Constructor/Parallel Mode/$$.

$childtable%
	cppad/local/parallel_ad.hpp%
	multi_thread/thread_test.cpp
%$$

$end