DNN Hardware Accelerator Bandwidth Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep neural networks (DNNs) face challenges in optimizing internal data transmission speed, which hinders processing speed acceleration.
Innovation Solution
A deep neural network hardware accelerator is designed with a network distributor that allocates individual bandwidths for different data types based on their proportions, and a processing element array that communicates data according to these allocated bandwidths, along with a bandwidth and utilization analysis unit that configures transmission bandwidths and controls multiplexers to prioritize data with larger sizes for efficient processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If uniform bandwidth is allocated to all data types, then device complexity is reduced, but data transmission efficiency deteriorates
Solution Approach 1:
The patent applies local quality by allocating different bandwidth proportions to different data types based on their specific characteristics. The network distributor analyzes data types and assigns customized bandwidth proportions (e.g., larger proportions for data types requiring faster transmission), rather than using a uniform allocation scheme. This resolves the contradiction by making the system more efficient through localized optimization while keeping the overall allocation mechanism relatively simple.
2Productivity
If bandwidth proportions are dynamically adjusted according to data type distribution, then data transmission efficiency is improved, but device complexity increases
Solution Approach 1:
The patent implements preliminary action by pre-configuring bandwidth proportions for different data types based on their characteristics. The system analyzes data type distribution in advance and establishes bandwidth allocation rules before actual data transmission occurs. This allows the system to efficiently handle different data types without requiring complex real-time dynamic adjustment mechanisms, thus improving transmission efficiency while controlling device complexity.
3Speed
If larger bandwidth is allocated to data types with larger sizes, then processing speed is improved, but use of energy increases
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting bandwidth allocation parameters based on data type characteristics and transmission requirements. The system monitors data type distribution and modifies bandwidth proportions accordingly, allocating larger bandwidths to data types that benefit most from faster transmission. This optimized parameter adjustment improves processing speed while minimizing energy consumption by avoiding unnecessary high-bandwidth allocation to all data types.
Data Source
AI summary
A DNN hardware accelerator and an operation method of the DNN hardware accelerator are provided. The DNN hardware accelerator includes: a network distributor for receiving an input data and distributing respective bandwidth of a plurality of data types of a target data amount based on a plurality of bandwidth ratios of the target data amount; and a processing element array coupled to the network distributor, for communicating data of the data types of the target data amount between the network distributor based on the distributed bandwidth of the data types.