Macrooperator organization of computing process

Abstract

The implementation of the computational process by using macro-operators based on multi-operand processing methods, the apparatus of chain shots and micro-operators with bit-by-bit pipeline computation is proposed. Such an organization has no disadvantages of known solutions, namely parallel forms, and therefore is an alternative for building new highly productive computer architectures. With such an organization, a bit-logarithmic representation of data is used, in which non-zero discharges of operands are encoded in the form of codes for the values of their weighting coefficients. The non-positive properties of such encoding allow independent parallel processing of operands of various mathematical expressions.

Thus, at the initial stage of calculations in a branch of the algorithm or the algorithm as a whole, a transformation is performed into a form determined by a multi-operator of multioperand processing, a chain fraction or combinations of the above. The new, resulting form is considered as a one-stage or multi-stage pipeline, in which calculations are performed simultaneously, without the rules of successive dependent transformations, thus overcoming the dependence on the data. The overall performance of the result determination increases to the performance of the pipeline step without the cost of overclocking it. Another measure of performance is the equivalent number of operations performed in a new form of representation of an algorithm or its branch.