\(\newcommand{\W}[1]{ \; #1 \; }\) \(\newcommand{\R}[1]{ {\rm #1} }\) \(\newcommand{\B}[1]{ {\bf #1} }\) \(\newcommand{\D}[2]{ \frac{\partial #1}{\partial #2} }\) \(\newcommand{\DD}[3]{ \frac{\partial^2 #1}{\partial #2 \partial #3} }\) \(\newcommand{\Dpow}[2]{ \frac{\partial^{#1}}{\partial {#2}^{#1}} }\) \(\newcommand{\dpow}[2]{ \frac{ {\rm d}^{#1}}{{\rm d}\, {#2}^{#1}} }\)
check_for_nan
Check an ADFun Object For Nan Results
Syntax
check_for_nan
( b )check_for_nan
()get_check_for_nan
( vec , file )Debugging
If NDEBUG
is not defined, and
the result of a forward or reverse
calculation contains a nan ,
CppAD can halt with an error message.
f
For the syntax where b is an argument, f has prototype
ADFun<
Base > f
(see ADFun<
Base > constructor ).
For the syntax where b is the result,
f has prototype
const ADFun<
Base > f
b
This argument or result has prototype
bool
b
If b is true (false),
future calls to f . Forward
will (will not) check for nan
.
Default
The value for this setting after construction of f is true. The value of this setting is not affected by calling Dependent for this function object.
Error Message
If this error is detected during zero order forward mode,
the values of the independent variables that resulted in the nan
are written to a temporary binary file.
This is so that you can run the original source code with those values
to see what is causing the nan
.
vector_size
The error message with contain the text
vector_size
= vector_size followed the newline character
'\n'
.
The value of vector_size is the number of elements
in the independent vector.
file_name
The error message with contain the text
file_name
= file_name followed the newline character
'\n'
.
The value of file_name is the name of the temporary file
that contains the dependent variable values.
index
The error message will contain the text
index
= index followed by the newline character '\n'
.
The value of index is the lowest dependent variable index
that has the value nan
.
get_check_for_nan
This routine can be used to get the independent variable
values that result in a nan
.
vec
This argument has prototype
CppAD::vector<
Base >& vec
It size must be equal to the corresponding value of
vector_size
in the corresponding error message.
The input value of its elements does not matter.
Upon return, it will contain the values for the independent variables,
in the corresponding call to Independent ,
that resulted in the nan
.
(Note that the call to Independent
uses an vector with elements
of type AD<
Base > and vec has elements of type
Base .)
file
This argument has prototype
const std::string&
file
It must be the value of file_name in the corresponding error message.
Example
The file check_for_nan.cpp contains an example and test of these operations.