Processing unit, related device and tensor operation method

Abstract

Provided are a processing unit, a related device and a tensor operation method. The processing unit includes: a plurality of computing units, forming a computing matrix with n rows and m columns, where n and m are non-zero natural numbers; a computing unit controller, configured to: The control calculation matrix works in the multicast data input mode. The data is broadcast to all the calculation units in the corresponding column by column, and to all the calculation units in the corresponding row by row. When the bandwidth of the external environment does not meet the predetermined bandwidth requirements, the control calculation matrix Working in the pulse data input mode, the computing unit receives data from the computing unit in the same row in the previous column and the computing unit in the previous row in the same column to support tensor operations. The embodiment of the present disclosure flexibly configures the working mode of the computing matrix according to the bandwidth of the external environment of the processing unit, so that the bandwidth of the external environment of the processing unit and the computing capability are adapted, and the computing energy efficiency of the processing unit is improved.

Description

technical field [0001] The present disclosure relates to the field of chips, and in particular, to a processing unit, a related device, and a tensor operation method. Background technique [0002] Deep learning is currently widely used in face recognition, speech recognition, automatic driving and other fields. Since deep learning relies on a large number of repeated tensor operations such as convolution and matrix operations, traditional hardware is inefficient in executing the corresponding algorithms. Therefore, a computing architecture dedicated to executing them has emerged as the times require. The deep learning processing units in these architectures employ computational matrices composed of multiple computational units. Each calculation unit in the calculation matrix performs convolution and operation of elements in the matrix operation, and then accumulates the operation results to obtain the tensor operation result. There are generally two ways to transmit the el...

AI Technical Summary

Problems solved by technology

When the bandwidth of the external environment is low, the multicast data input mode cannot be used for efficient calculation, and the idle waiting state of the computing unit is often

When the bandwidth of the external environment is high, the pulse data input mode cannot fully utilize the bandwidth and resources

These all cause the external environmental bandwidth and computing power of the processing unit to be unsuitable, reducing the computing energy efficiency of the processing unit

Images