grad
grad |
Purpose
Evaluate the gradiant of a constraint or objective on the provided assignment of the decision variables.
Synopsis
int grad(XPRMcontext ctx, XPRMnlpdata nlpdata, int nb, double *grad, const double *x, double *val, int dense);
Arguments
|
ctx
|
Mosel's execution context
|
|
nlpd
|
Problem information as generated by
loadprob
|
|
nb
|
Constraint number to evaluate or
NLEVAL_OBJ for the opbjective,
NLEVAL_ALL for everything,
NLEVAL_NL for nonlinear constraints or
NLEVAL_LIN linear constraints.
|
|
grad
|
Reference that receives the gradiant. In case of multiple evaluations, this must be an array large enough to store all the results
|
|
x
|
Values assigned to the decision variables
|
|
val
|
Reference that receives the result of the evaluation (or
NULL). In case of multiple evaluations, this must be an array large enough to store all the results
|
|
dense
|
When processing a single constraint, return data in dense form
|
Return value
0 if successful or 1 in case of error
Further information
1. This function computes the gradient and evaluation of the objective or constraints. The
val argument is handled like with function
eval.
2. The
dense option is used only when processing a single constraint or the objective: the result is returned in the form of a dense array of coefficients, the
grad array must be of size
nlpdata->nbvar. In all other cases the result is returned in the form of a sparse array: for the objective the size of the array must be
nlpdata->nzrobj and the variable references have to be obtained from
nlpdata->vrefobj. For a constraint
c the size of the array is
nlpdata->ctrstart[c+1]-nlpdata->ctrstart[c] and the variable references are taken from
nlpdata->vref[nlpdata->ctrstart[c]] to
nlpdata->vref[nlpdata->ctrstart[c+1]].
3. If the function is used to process all constraints (
NLEVAL_ALL) the result array is of size
npldata->nzr, for linear constraints only (
NLEVAL_LIN) it is
npldata->nzr_lin and for the nonlinear constraints (
NLEVAL_NL) it is
npldata->nzr-npldata->nzr_lin.