Optimizing output vector data generation using a formatted matrix data structure

摘要

A computer system retrieves a packet that includes non-zero elements that correspond to sparse-matrix rows. Within the packet, the non-zero elements are stored in predefined fields that each correspond to one of the sparse-matrix rows. The computer system computes output values to correspond with each of the sparse-matrix rows using the non-zero elements and corresponding input values. In turn, the computer system stores the computed output values in consecutive locations within an output buffer and processes the output values accordingly.

主权项

1. A computer-implemented method comprising:
retrieving, by one or more processors, a first packet that includes an output offset and one or more non-zero elements each corresponding to a different one of a plurality of rows in a sparse-matrix, wherein each of the one or more non-zero elements is stored in one or more of a plurality of predefined fields included in the first packet that each correspond to one of the plurality of rows; computing a plurality of output values to correspond with each of the plurality of rows using one or more of the non-zero elements and one or more corresponding input values included in an input vector; processing the plurality of output values; and performing a single block write operation to store the plurality of output values in the plurality of consecutive locations using the retrieved output offset.