本文整理汇总了C++中CppAD::NumArg方法的典型用法代码示例。如果您正苦于以下问题:C++ CppAD::NumArg方法的具体用法?C++ CppAD::NumArg怎么用?C++ CppAD::NumArg使用的例子?那么, 这里精选的方法代码示例或许可以为您提供帮助。您也可以进一步了解该方法所在类CppAD
的用法示例。
在下文中一共展示了CppAD::NumArg方法的9个代码示例,这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞,您的评价将有助于系统推荐出更棒的C++代码示例。
示例1: reverse_next
/*!
Fetch the next operator during a reverse sweep.
Use reverse_start to initialize to reverse play back.
The first call to reverse_next (after reverse_start) will give the
last operator in the recording.
We use the notation reverse_routine to denote the set
reverse_start, reverse_next, reverse_csum, reverse_cskip.
\param op [in,out]
The input value of \c op must be its output value from the
previous call to a reverse_routine.
Its output value is the next operator in the recording (in reverse order).
The last operator sets op equal to EndOp.
\param op_arg [in,out]
The input value of \c op_arg must be its output value from the
previous call to a reverse_routine.
Its output value is the
beginning of the vector of argument indices for this operation.
The last operator sets op_arg equal to the beginning of the
argument indices for the entire recording.
For speed, \c reverse_next does not check for the special cases
<tt>op == CSumOp</tt> or <tt>op == CSkipOp</tt>. In these cases, the other
return values from \c reverse_next must be corrected by a call to
\c reverse_csum or \c reverse_cskip respectively.
\param op_index [in,out]
The input value of \c op_index must be its output value from the
previous call to a reverse_routine.
Its output value
is the index of this operator in the recording. Thus the output
value following the previous call to reverse_start is equal to
the number of variables in the recording minus one.
In addition, the output value decreases by one with each call to
reverse_next.
The last operator sets op_index equal to 0.
\param var_index [in,out]
The input value of \c var_index must be its output value from the
previous call to a reverse_routine.
Its output value is the
index of the primary (last) result corresponding to the operator op.
The last operator sets var_index equal to 0 (corresponding to BeginOp
at beginning of operation sequence).
*/
void reverse_next(
OpCode& op, const addr_t*& op_arg, size_t& op_index, size_t& var_index)
{ using CppAD::NumRes;
using CppAD::NumArg;
CPPAD_ASSERT_UNKNOWN( op_ == op );
CPPAD_ASSERT_UNKNOWN( op_arg == op_arg_ );
CPPAD_ASSERT_UNKNOWN( op_index == op_index_ );
CPPAD_ASSERT_UNKNOWN( var_index == var_index_ );
// index of the last result for the next operator
CPPAD_ASSERT_UNKNOWN( var_index_ >= NumRes(op_) );
var_index = var_index_ -= NumRes(op_);
// next operator
CPPAD_ASSERT_UNKNOWN( op_index_ > 0 );
op_index = --op_index_; // index
op = op_ = OpCode( op_rec_[ op_index_ ] ); // value
// first argument for next operator
op_arg = op_arg_ -= NumArg(op);
CPPAD_ASSERT_UNKNOWN( op_arg_rec_.data() <= op_arg_ );
CPPAD_ASSERT_UNKNOWN(
op_arg_ + NumArg(op) <= op_arg_rec_.data() + op_arg_rec_.size()
);
}
示例2: forward_cskip
/*!
Correct \c forward_next return values when <tt>op == CSkipOp</tt>.
\param op [in]
The input value of op must be the return value from the previous
call to \c forward_next and must be \c CSkipOp. It is not modified.
\param op_arg [in,out]
The input value of \c op_arg must be the return value from the
previous call to \c forward_next. Its output value is the
beginning of the vector of argument indices for the next operation.
\param op_index [in]
The input value of \c op_index does must be the return value from the
previous call to \c forward_next. Its is not modified.
\param var_index [in,out]
The input value of \c var_index must be the return value from the
previous call to \c forward_next. It is not modified.
*/
void forward_cskip(
OpCode& op, const addr_t*& op_arg, size_t& op_index, size_t& var_index)
{ using CppAD::NumRes;
using CppAD::NumArg;
CPPAD_ASSERT_UNKNOWN( op_ == op );
CPPAD_ASSERT_UNKNOWN( op_arg == op_arg_ );
CPPAD_ASSERT_UNKNOWN( op_index == op_index_ );
CPPAD_ASSERT_UNKNOWN( var_index == var_index_ );
CPPAD_ASSERT_UNKNOWN( op == CSkipOp );
CPPAD_ASSERT_UNKNOWN( NumArg(CSkipOp) == 0 );
CPPAD_ASSERT_UNKNOWN(
op_arg[4] + op_arg[5] == op_arg[ 6 + op_arg[4] + op_arg[5] ]
);
/*
The only thing that really needs fixing is op_arg_.
Actual number of arugments for this operator is
7 + op_arg[4] + op_arg[5]
We must change op_arg_ so that when you add NumArg(CSkipOp)
you get first argument for next operator in sequence.
*/
op_arg = op_arg_ += 7 + op_arg[4] + op_arg[5];
CPPAD_ASSERT_UNKNOWN( op_arg_rec_.data() <= op_arg_ );
CPPAD_ASSERT_UNKNOWN(
op_arg_ + NumArg(op) <= op_arg_rec_.data() + op_arg_rec_.size()
);
CPPAD_ASSERT_UNKNOWN( var_index_ < num_var_rec_ );
}
示例3: primary
/*!
Fetch the next operator during a forward sweep.
Use forward_start to initialize to the first operator; i.e.,
the BeginOp at the beginning of the recording.
We use the notation forward_routine to denote the set
forward_start, forward_next, forward_csum, forward_cskip.
\param op [in,out]
The input value of \c op must be its output value from the
previous call to a forward_routine.
Its output value is the next operator in the recording.
For speed, \c forward_next does not check for the special cases
where <tt>op == CSumOp</tt> or <tt>op == CSkipOp</tt>. In these cases,
the other return values from \c forward_next must be corrected by a call
to \c forward_csum or \c forward_cskip respectively.
\param op_arg [in,out]
The input value of \c op_arg must be its output value form the
previous call to a forward routine.
Its output value is the
beginning of the vector of argument indices for this operation.
\param op_index [in,out]
The input value of \c op_index must be its output value form the
previous call to a forward routine.
Its output value is the index of the next operator in the recording.
Thus the ouput value following the previous call to forward_start is one.
In addition,
the output value increases by one with each call to forward_next.
\param var_index [in,out]
The input value of \c var_index must be its output value form the
previous call to a forward routine.
Its output value is the
index of the primary (last) result corresponding to the operator op.
*/
void forward_next(
OpCode& op, const addr_t*& op_arg, size_t& op_index, size_t& var_index)
{ using CppAD::NumRes;
using CppAD::NumArg;
CPPAD_ASSERT_UNKNOWN( op_ == op );
CPPAD_ASSERT_UNKNOWN( op_arg == op_arg_ );
CPPAD_ASSERT_UNKNOWN( op_index == op_index_ );
CPPAD_ASSERT_UNKNOWN( var_index == var_index_ );
// index for the next operator
op_index = ++op_index_;
// first argument for next operator
op_arg = op_arg_ += NumArg(op_);
// next operator
op = op_ = OpCode( op_rec_[ op_index_ ] );
// index for last result for next operator
var_index = var_index_ += NumRes(op);
CPPAD_ASSERT_UNKNOWN( op_arg_rec_.data() <= op_arg_ );
CPPAD_ASSERT_UNKNOWN(
op_arg_ + NumArg(op) <= op_arg_rec_.data() + op_arg_rec_.size()
);
CPPAD_ASSERT_UNKNOWN( var_index_ < num_var_rec_ );
}
示例4: reverse_cskip
void reverse_cskip(
OpCode& op, const addr_t*& op_arg, size_t& op_index, size_t& var_index)
{ using CppAD::NumRes;
using CppAD::NumArg;
CPPAD_ASSERT_UNKNOWN( op_ == op );
CPPAD_ASSERT_UNKNOWN( op_arg == op_arg_ );
CPPAD_ASSERT_UNKNOWN( op_index == op_index_ );
CPPAD_ASSERT_UNKNOWN( var_index == var_index_ );
CPPAD_ASSERT_UNKNOWN( op == CSkipOp );
CPPAD_ASSERT_UNKNOWN( NumArg(CSkipOp) == 0 );
/*
The variables that need fixing are op_arg_ and op_arg. Currently,
op_arg points first arugment for the previous operator.
*/
--op_arg;
op_arg = op_arg_ -= (op_arg[0] + 4);
CPPAD_ASSERT_UNKNOWN(
op_arg[1] + op_arg[2] == op_arg[ 3 + op_arg[1] + op_arg[2] ]
);
CPPAD_ASSERT_UNKNOWN( op_index_ < op_rec_.size() );
CPPAD_ASSERT_UNKNOWN( op_arg_rec_.data() <= op_arg_ );
CPPAD_ASSERT_UNKNOWN( var_index_ < num_var_rec_ );
}
示例5: reverse_csum
void reverse_csum(
OpCode& op, const addr_t*& op_arg, size_t& op_index, size_t& var_index)
{ using CppAD::NumRes;
using CppAD::NumArg;
CPPAD_ASSERT_UNKNOWN( op_ == op );
CPPAD_ASSERT_UNKNOWN( op_arg == op_arg_ );
CPPAD_ASSERT_UNKNOWN( op_index == op_index_ );
CPPAD_ASSERT_UNKNOWN( var_index == var_index_ );
CPPAD_ASSERT_UNKNOWN( op == CSumOp );
CPPAD_ASSERT_UNKNOWN( NumArg(CSumOp) == 0 );
/*
The variables that need fixing are op_arg_ and op_arg. Currently,
op_arg points to the last argument for the previous operator.
*/
// last argument for this csum operation
--op_arg;
// first argument for this csum operation
op_arg = op_arg_ -= (op_arg[0] + 4);
// now op_arg points to the first argument for this csum operator
CPPAD_ASSERT_UNKNOWN(
op_arg[0] + op_arg[1] == op_arg[ 3 + op_arg[0] + op_arg[1] ]
);
CPPAD_ASSERT_UNKNOWN( op_index_ < op_rec_.size() );
CPPAD_ASSERT_UNKNOWN( op_arg_rec_.data() <= op_arg_ );
CPPAD_ASSERT_UNKNOWN( var_index_ < num_var_rec_ );
}
示例6: next_reverse
void next_reverse(
OpCode& op, const addr_t*& op_arg, size_t& op_index, size_t& var_index)
{ using CppAD::NumRes;
using CppAD::NumArg;
// index of the last result for the next operator
CPPAD_ASSERT_UNKNOWN( var_index_ >= NumRes(op_) );
var_index = var_index_ -= NumRes(op_);
// next operator
CPPAD_ASSERT_UNKNOWN( op_index_ > 0 );
op_index = --op_index_; // index
op = op_ = OpCode( rec_op_[ op_index_ ] ); // value
// first argument for next operator
CPPAD_ASSERT_UNKNOWN( op_arg_ >= NumArg(op) );
op_arg_ -= NumArg(op); // index
op_arg = op_arg_ + rec_op_arg_.data(); // pointer
}
示例7: reverse_csum
void reverse_csum(
OpCode& op, const addr_t*& op_arg, size_t& op_index, size_t& var_index)
{ using CppAD::NumRes;
using CppAD::NumArg;
CPPAD_ASSERT_UNKNOWN( op == CSumOp );
CPPAD_ASSERT_UNKNOWN( NumArg(CSumOp) == 0 );
/*
The things needs fixing are op_arg_ and op_arg. Currently,
op_arg points first arugment for the previous operator.
*/
--op_arg;
op_arg_ -= (op_arg[0] + 4);
op_arg = op_arg_ + rec_op_arg_.data();
CPPAD_ASSERT_UNKNOWN(
op_arg[0] + op_arg[1] == op_arg[ 3 + op_arg[0] + op_arg[1] ]
);
CPPAD_ASSERT_UNKNOWN( op_index_ < rec_op_.size() );
CPPAD_ASSERT_UNKNOWN( op_arg_ + NumArg(op) <= rec_op_arg_.size() );
CPPAD_ASSERT_UNKNOWN( var_index_ < num_rec_var_ );
}
示例8: primary
/*!
Correct \c next_forward return values when <tt>op == CSumOp</tt>.
\param op
The input value of op must be the return value from the previous
call to \c next_forward and must be \c CSumOp.
\param op_arg
The input value of *op_arg must be the return value from the
previous call to \c next_forward. Its output value is the
beginning of the vector of argument indices for this operation.
\param op_index
The input value of op_index does must be the return value from the
previous call to \c next_forward. Its output value
is the index of this operator in the recording.
\param var_index
The input value of var_index must be the return value from the
previous call to \c next_forward. Its output value is the
index of the primary (last) result corresponding to this.
*/
void forward_csum(
OpCode& op, const addr_t*& op_arg, size_t& op_index, size_t& var_index)
{ using CppAD::NumRes;
using CppAD::NumArg;
CPPAD_ASSERT_UNKNOWN( op == CSumOp );
CPPAD_ASSERT_UNKNOWN( NumArg(CSumOp) == 0 );
CPPAD_ASSERT_UNKNOWN(
op_arg[0] + op_arg[1] == op_arg[ 3 + op_arg[0] + op_arg[1] ]
);
/*
The only thing that really needs fixing is op_arg_.
Actual number of arugments for this operator is
op_arg[0] + op_arg[1] + 4.
We must change op_arg_ so that when you add NumArg(CSumOp)
you get first argument for next operator in sequence.
*/
op_arg_ += op_arg[0] + op_arg[1] + 4;
CPPAD_ASSERT_UNKNOWN( op_arg_ + NumArg(op) <= rec_op_arg_.size() );
CPPAD_ASSERT_UNKNOWN( var_index_ < num_rec_var_ );
}
示例9: next_forward
void next_forward(
OpCode& op, const addr_t*& op_arg, size_t& op_index, size_t& var_index)
{ using CppAD::NumRes;
using CppAD::NumArg;
// index for the next operator
op_index = ++op_index_;
// first argument for next operator
op_arg_ += NumArg(op_); // index
op_arg = op_arg_ + rec_op_arg_.data(); // pointer
// next operator
op = op_ = OpCode( rec_op_[ op_index_ ] );
// index for last result for next operator
var_index = var_index_ += NumRes(op);
CPPAD_ASSERT_UNKNOWN( op_arg_ + NumArg(op) <= rec_op_arg_.size() );
CPPAD_ASSERT_UNKNOWN( var_index_ < num_rec_var_ );
}