An OPU instruction set definition method for CNN acceleration

Abstract

The invention discloses an OPU instruction set defining method for CNN acceleration, and relates to the field of instructions of CNN acceleration processors, and the method comprises the steps of defining a conditional instruction, defining an unconditional instruction and setting the instruction granularity, wherein the unconditional instruction provides configuration parameters for the conditional instruction, the conditional instruction sets a trigger condition, the trigger condition is hard-written in hardware, the conditional instruction sets a trigger condition register corresponding tothe conditional instruction, the conditional instruction is executed after the trigger condition is met, the unconditional instruction is directly executed after being read, and the content of the parameter register is replaced. According to the CNN network and acceleration requirements, the calculation modes of the parallel input and output channels are selected, and the instruction granularity is set. The instruction set provided by the invention avoids the problem that the order of the instructions cannot be predicted due to large uncertainty of the operation period. The instruction set andthe corresponding processor OPU can be realized by an FPGA or an ASIC. The OPU can accelerate different target CNN networks, and hardware reconstruction is avoided.

Description

technical field [0001] The invention relates to the field of a method for defining a CNN accelerator instruction set, in particular to a method for defining an OPU instruction set for CNN acceleration. Background technique [0002] Deep convolutional neural networks (CNNs) have demonstrated high accuracy in various applications, such as visual object recognition, speech recognition, and object detection, etc. However, its breakthrough in accuracy comes at the cost of high computational cost, which needs to be accelerated by computing clusters, GPUs and FPGAs. Among them, the FPGA accelerator has the advantages of high energy efficiency, good flexibility, and strong computing power, especially in deep CNN applications on edge devices such as speech recognition and visual object recognition on smartphones; it usually involves architecture exploration and optimization , RTL programming, hardware implementation and software-hardware interface development. With the development, ...

AI Technical Summary

Problems solved by technology

On the other hand, when it comes to external memory usage, the cyclic simulation of memory read and write operations is not very accurate, because additional refresh time and other overhead may occur during external memory usage; if the instruction is executed immediately after decoding, Then the order of operations can only be controlled by the order of the instruction sequence; if the operation cycle is not accurately simulated, it will become difficult to control the starting point of the operations executed in parallel; at the same time, the starting conditions of the main business have limited changes and are usually reached in the first few steps It is triggered after a certain state, resulting in a large uncertainty in the execution time of instructions. Therefore, an instruction set definition method is needed to define instructions to overcome the above problems, provide OPU instruction sets to reorganize network mappings of different structures into specific structures, and optimize instruction control. The universality of the processor can realize the configuration of different target networks according to the instructions, and realize the general CNN acceleration through the OPU

Images