Subroutine for numerical minimization of function of many parameters using analytically evaluated gradient

A program of numerical minimization of function of many parameters using analytically evaluated gradient is carried out. There are several algorithms of minimization implemented in this program: Golden section method for one parameter minimization, simplex method, variable metrics method, conjugate directions method, Newton's method, try and fail method, random search. It is possible to combine different methods. For the chosen combination of methods there is possible a multiple minimization procedure from different start points. For those algorithms which use function gradient there is used a user supplied subroutine for gradient evaluation (supposed that exact analytical formulas are used). Program language is Fortran 77. Total size of source code is about 3300 rows