Abstract: A system and method for performing neural network calculations may include selecting a size in bits for representing a plurality of weight elements of the neural network based on a value of the weight elements. In each computational cycle: if the size in bits of a weight element of the plurality of weight elements is N, configuring an N*K multiply accumulator to perform one multiply-accumulate operation of a K-bit data element and the N-bit weight element; and if the size in bits of at least two N/M-bit weight elements of the plurality of weight elements is N/M, configuring the N*K multiply accumulator to perform up to N/M multiply-accumulate operations, each of a K-bit, data element and an N/M-bit weight element, where N, K and M are integers bigger than one, N is a power of 2, M is even and N?M.

Abstract: A system and method of data transfer that collapses a multi-dimensional data array while preserving neighboring connectivity. One or more program instructions may be received that request input data comprising a sub-set of data elements from a first data array in an external memory. The sub-set of data elements may be transferred from the first data array in the external memory to a second data array in an internal memory. The retrieved sub-set of requested data elements may be stored in the second array in respective positions translated from the positions in the first array by a plurality of omitted rows and columns from the first data array from which no data elements are requested.

Abstract: A system and method of data transfer that collapses a multi-dimensional data array while preserving neighboring connectivity. One or more program instructions may be received that request input data comprising a sub-set of data elements from a first data array in an external memory. The sub-set of data elements may be transferred from the first data array in the external memory to a second data array in an internal memory. The retrieved sub-set of requested data elements may be stored in the second array in respective positions translated from the positions in the first array by a plurality of omitted rows and columns from the first data array from which no data elements are requested.