FIELD: information technology.
SUBSTANCE: method of performing floating-point operations with a floating-point processor which is configured to perform floating-point operations, having a first accuracy, involves the following steps: selection of sub-accuracy which is less than the first accuracy for performing floating-point operations over one or more floating-point numbers, where the selection of sub-accuracy results in one or more redundancy bits for each of the said one or more floating-point numbers; cutting off power to logic circuits which could have been used to process one or more redundancy bits; and performing floating-point operations while power to logic circuits is cut off.
EFFECT: low power consumed by the processor.
24 cl, 4 dwg
