MASSIVELY PARALLEL NEURAL INFERENCE DATA PROCESSING ELEMENTS

摘要

Massively parallel neural inference data processing elements are provided. A plurality of multipliers are arranged in a plurality of groups of equal size. Each of the plurality of multipliers is configured to apply weighting to input activation in parallel to produce an output. A plurality of adders are operatively connected to one of the groups of multipliers. Each of the plurality of adders is designed to add the outputs of the multipliers within the group associated with them in parallel in order to generate a partial sum. A plurality of function blocks are operatively connected to one of the plurality of adders. Each of the plurality of function blocks is designed to apply a function to the partial sum of the adder associated with it in parallel to generate an output value.