Neural Network Processor Exception Handling via Instruction Queue Reordering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The neural network processor's reliance on an interrupt mechanism for exception processing leads to significant time delays, thereby reducing the data processing efficiency of the entire neural network hardware system.
Innovation Solution
An instruction execution method that involves acquiring exceptional signals, determining corresponding exception processing instructions, and generating a second instruction queue to be executed by the neural network processor, allowing for timely error processing and improved data processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If an interrupt mechanism is used for exception processing in the neural network processor, then the exception can be handled, but the data processing efficiency is reduced due to time delays
Solution Approach 1:
The patent applies preliminary action by pre-allocating a dedicated exception processing resource (second processor) that is ready to handle exceptions immediately when they occur. Instead of using an interrupt mechanism that pauses normal processing, the system has a standby processor that can take over exception handling without delaying the main data processing flow, thus resolving the contradiction between reliable exception handling and high data processing efficiency
Solution Approach 2:
The patent segments the processing functions by separating exception handling from normal data processing. The first processor handles normal neural network computations while the second processor is dedicated solely to exception handling. This segmentation allows both functions to operate independently and simultaneously, eliminating the time delays caused by interrupt mechanisms while maintaining both reliability and productivity
2Productivity
If the neural network processor continuously processes large amounts of data, then data processing capacity is high, but any exception requires immediate processing which interrupts the flow and reduces efficiency
Solution Approach 1:
The patent introduces an intermediary mechanism (the second processor) that mediates between the data processing flow and exception handling requirements. When an exception occurs in the first processor, the second processor acts as an intermediary that can immediately begin handling the exception without interrupting the continuous data processing flow of the first processor, thus eliminating time loss while maintaining high processing capacity
Solution Approach 2:
The patent ensures continuity of useful action by maintaining independent operation of both processors. The first processor continues its data processing tasks without interruption while the second processor handles exceptions in parallel. This continuous operation of both processors simultaneously eliminates exception processing delays while preserving high data processing capacity
Data Source
AI summary
The present disclosure provides an instruction execution method, device, and electronic equipment. In the instruction execution method described above, after obtaining an exceptional signal generated by a neural network processor during an operation, the electronic equipment determines an exception processing instruction corresponding to the exceptional signal according to the exceptional signal, then it determines a first instruction queue needed to be executed by the neural network processor, and then it generates a second instruction queue based on the exception processing instruction and the first instruction queue, and finally it controls the neural network processor to execute the second instruction queue, so that errors encountered by the neural network processor can be timely processed, thereby shortening the error processing delay and improving the data processing efficiency of the hardware system in the electronic equipment.