Chip comprising multiply-accumulate module, terminal and control method

Abstract

The invention discloses a chip comprising a multiply-accumulate module, a terminal and a control method, and relates to the field of chips. The chip comprises a multiplication and accumulation module,and the multiplication and accumulation module comprises a fixed point general unit, a floating point special unit and an output selection unit. The fixed point general unit and the floating point special unit are respectively connected with the output selection unit; and the fixed-point general unit is also connected with the floating point special unit, and the fixed-point general unit and thefloating point special unit share a group of multipliers. In the multiplication and accumulation module of the chip, fixed-point operation and floating-point operation are integrated on a circuit, sothat the multiplication and accumulation module realizes fixed-point operation in the circuit, and floating-point operation can be realized at the same time; the fixed-point operation unit and the floating-point operation unit share the multiplier, so that the total number of used devices is reduced, and the occupied area of the fixed-point operation unit and the floating-point operation unit on achip and the power consumption during operation are reduced.

Description

technical field [0001] The present application relates to the field of chips, in particular to a chip including a multiply-accumulate module, a terminal and a control method. Background technique [0002] The multiply-accumulate module is a basic calculation module on the chip, which is widely used in such as central processing unit (Central Processing Unit, CPU), field-programmable gate array (Field-Programmable Gate Array, FPGA), application specific integrated circuit (Application Specific Integrated Circuits) , ASIC), graphics processing unit (Graphics Processing Unit, GPU) or other artificial intelligence (Artificial Intelligence, AI) chips and other chips. [0003] Taking the chip used for neural network model calculation as an example, there are two kinds of multiply-accumulate modules on the chip: the first multiply-accumulate module for fixed-point operations (also called integer operations), and the first for floating-point operations. Two multiply-accumulate modu...

AI Technical Summary

Problems solved by technology

[0004] Due to the need to implement two multiply-accumulate modules on the chip at the same time, the chip area and power consumption of the chip are relatively large

Images