Previous: Dynamic Data Structures, Up: Storage Management
One final aspect of these storage schemes must be considered — when should they be used? It is clear that for new work in CONGEN, C should be used for programming because it provides all these facilities in a straightforward manner. (CHARMM and CONGEN would have been more advanced had C be used from the beginning.) However, if new Fortran code must be integrated or maintained, then these schemes are applicable. The stack and the data structure storage on the heap were developed primarily to deal with an enormous number of arrays whose sizes must be able to change. The other design criteria came later. It should be obvious that one pays a price for these schemes, and one must determine whether the benefits are worth the cost.
One disadvantage of these methods is that they are unfamiliar, and therefore, they take time to learn. Unfortunately, a subjective judgment is required to estimate how large a disadvantage this is. In my opinion, this chapter will enable an interested person to learn this material in a day or two. The subroutines to use these schemes are sufficiently comprehensive to permit effective use of these schemes without additional programming. Further, the ideas presented in this document are worth learning as they are valuable outside the context of CONGEN. Besides its application for local storage, the stack is useful for problems which require backtracking or which can be solved recursively. The heap is useful for problems which involve management of complex lists.
A second disadvantage of these schemes is that they are more cumbersome than the techniques they replace. There is no arguing this point. However, when used with the conventions described here, clarity is not sacrificed.
Let us now consider the advantages. We shall confine ourselves to the analysis facility where they are used extensively.
First, as the size of a system being analyzed grows to the limits of what is supported, one need only change the size of the stack and the heap to accommodate a bigger system. When a subroutine has as many arrays as the analysis facility has, this is vital.
Next, although references to parts of a data structure are more cumbersome, referencing a whole structure is much more concise. The subroutines in the analysis facility illustrates the compactness of the subroutine call despite the large amount of information that is being passed. Further, the inclusion of the comparison data structures is trivial; we merely have to create another set of base and length arrays.
Thirdly, the data structures can be referred to as a unit. This aids the job of programming because now we can deal with bigger units of information without having to worry about the details. These data structures on the heap allow a programmer to ignore their contents until they are ready to be used. This can simplify the task of modifying a program the size of CONGEN.