A computer architecture which is centered around the transformation of expressions facilitates the generation of programs from definitions, axioms, and theorems. It is shown that some of the well known methods to construct correct programs are directly available on this novel computer. Due to its novel architecture the need for support by large software systems and entailing systems size is diminished or has disappeared. Using this novel computer employing a lambda reduction language as machine language it is demonstrated how suitably chosen invariants get transformed into an executable program which is correct by construction. The various types of information which get incorporated into the result, as well as those parts of the original expression denoting the invariant which get expelled because they are constant, are discussed. It is demonstrated why the above mentioned novel computer architecture frees the user from being concerned with inessential problems, while on the other hand the use of higher level programming languages introduces additional problems due to the semantics of these languages. They necessitate considerations alien to solving the original problem. Crossreferences to other disciplines are pointed out which employ similar concepts.

This paper is to present the basic concept of a hardwired storage management system. Information about the status of the storage cells is held in two separate bit-planes whose contents are interrogated and updated by a special processor. The system processes requests for the releases of memory space, returning and receiving base addresses, respectively. It operates concurrently to the CPU, thereby relieving the latter of a considerable workload. The hardware of the processor is designed to accommodate a fast string search algorithm that employs ‘don't care’ states for blockwise addressing of the bit-planes.

This paper describes a particular realization of a Lambda-Red language as a machine language. Parts of it resemble the Lambda Calculus or a transposed form of it. A constructor syntax is employed such that a linearized preorder representation of the syntax tree is the information structure on which the machine operates. Instances of reduction rules are recognized by combinations of constructors and atoms. Reduction rules with 1, 2 and 3 constructors and/or atoms have been described.A recursive control structure forms the essential part of the implementation. There is a one to one correspondence between constructor syntax and control structure rather than a simulation of recursive structure by von Neumann type instruction sequencing. Thus the system is easily expandable by new constructors and atoms. Execution, that is the application of reduction rules, is subsumed under editing. Emphasis has been on the following pragmatic concepts: locality of action, directness, and security. There are no error messages, errors appear as irreducible expressions. There is no "RUN" instruction, instead the number of reductions to be performed is specified. Whatever happens - it is restricted to subexpressions. The user has the security that all his actions are limited and predictable.The machine language resembles a higher level programming language, but it is "variable-free": expressions are named, not boxes which contain expressions. There seems to be a concept of late binding. However, this is not made a matter of principle, but a matter of degree. Between a general statement of a problem and the result are many intermediate representations - all in the same language - corresponding to various degrees of binding parameters. The user has complete freedom in the choice of variable names. There is a specially developped protection system which avoids confusion of variables.The efficiency of the system is limited by the rate characters can be processed which is essentially determined by the control store cycletime. Measurements will be performed as soon as a simulation of the system is available. Three other important subjects, namely arithmetic, the definition of constants, and source sink input-output will be reported later.

The expenditure of resources to provide via software more structure and more function in conventional machines has become excessive. The answer to this problem may consist in providing more structure and more function in hardware. With respect to structure, we describe 1) the binary tree as basic data and control structure, 2) the implementation of a binary tree by a conventional memory, and 3) the associated hardware. The most important enhancement with respect to function, the newly introduced unit of action, deals with the handling of variables and substitution.