Computing system and method based on sram in-memory computing array

Abstract

The invention relates to a computing system and method based on an sram in-memory computing array. In the system, a data address generator is connected with a computing array in an sram memory; the data address generator outputs an address corresponding to the booth coding operation according to the multiplicand; the sram in-memory calculation array is used for determining a partial product according to a multiplicand written by an address; each column of the calculation array in the sram memory is correspondingly connected with one capacitor; each group is connected with an MUX device and an analog-to-digital converter in sequence; the MUX device combines the set number of paths of results of the set number of capacitors in each group into one path of result; the analog-to-digital converter performs analog-to-digital conversion on one path of a splicing multiplication module is connected with the plurality of analog-to-digital converters; and the splicing multiplication module performs splicing multiplication on a plurality of paths of results after analog-to-digital conversion to determine a calculation result. According to the invention, power consumption and time delay generated by coding operation can be reduced.

Description

technical field [0001] The invention relates to the field of sram memory computing arrays, in particular to a computing system and method based on sram memory computing arrays. Background technique [0002] As the second-generation neural network, convolutional neural network is currently widely used in face recognition, speech recognition and other fields. Booth encoding is a multiplication encoding method, and because the convolutional neural network often reuses the same multiplier many times, resulting in unnecessary delay and power consumption for multiple encoding operations; the traditional multiplier is performing A calculation requires a longer delay and consumes more energy during data transmission. [0003] Therefore, there is an urgent need for a computing system or method that reduces power consumption and delay generated by encoding operations. Contents of the invention [0004] The purpose of the present invention is to provide a computing system and metho...

AI Technical Summary

Problems solved by technology

Booth encoding is a multiplication encoding method, and because the same multiplier is often used repeatedly in the convolutional neural network, resulting in unnecessary delay and power consumption for multiple encoding operations; the traditional multiplier is performing A calculation requires a long delay and consumes more energy during data transmission

Images