to distinguish between these two types of decisions, derived decision

offsets are incremented by 10,000. Since decision offsets will never

be that large, large offsets are easily recognized as derived and are

adjusted to their true value when necessary.

5.1.2 Function-Variable Incidence Matrix Representation

 It is necessary that the Function-Variable Incidence Matrix be

stored in a data structure which allows for easy access by the algo-

rithms. Since the FVIM is a type of incidence matrix it is reasonable

to store it in the data structure used to store incidence matrices. The

incidence matrices are stored in a data structure called the SIM data

structure (Cunningham, 1973). Appendix A contains a description of the

SIM data structure. The principal difference between the normal

incidence matrix and the FVIM is the fact that an element in the FVIM is

actually a list of variable indices. To allow for that it is necessary

to modify the element entry in the normal SIM structure to accommodate an

extra word. This is easily done since the length of an entry is

variable in the SIM data structure. This extra word serves as a pointer

to the vector "VILIST." Thus, each FVIM entry has associated with it a

pointer to VILIST. The structure of VILIST is illustrated in Fig. 5-5.

Each FVIM element pointer points to a different VILIST cell. The VILIST

cells are merely lists of pointers to variable index definition cells.

This results in each element in an FVIM having associated with it a list

of variable indices, which is what is required.


5.2 A Versatile Memory System

 A part of the design philosophy of the GENDER system has been to

construct the different data structures necessary in separate labeled

COMM0N areas. For example, a special data structure was developed for