Data Processing Apparatus with Unified Control for Neural Network Operations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data processing systems face challenges in managing the diversification of recognition processing operations, leading to increased program and circuit sizes due to the complexity of neural network operations, particularly in handling both multiply-accumulate and post-processing portions efficiently.
Innovation Solution
A data processing apparatus is configured with a multiply-accumulate operation unit, an arithmetic operation result retention unit, a command processing unit, and a distribution unit that receives a combined parameter sequence to manage both multiply-accumulate and post-processing operations, allowing for flexible sequencing and reduced program/circuit size by interconnecting parameters for these operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the hardware is made compatible with a greater diversity of recognition processing operations, then the adaptability is improved, but the program size or circuit size increases
Solution Approach 1:
The patent implements a universal control mechanism where a single control unit can execute multiple types of recognition processing operations by dynamically loading different parameter sequences. The apparatus is designed to perform both multiply-accumulate operations and post-processing operations through a unified control structure that accepts external parameter sequences, allowing one hardware platform to serve multiple recognition tasks without requiring separate dedicated circuits for each operation type.
Solution Approach 2:
The patent employs dynamic parameter sequence loading to enable the control unit to adapt its behavior based on the specific recognition task at hand. Instead of static hardwired logic for each operation type, the system dynamically loads parameter sequences that define the operational characteristics, allowing the same hardware to flexibly switch between different recognition processing operations by changing the parameter set rather than the physical circuit configuration.
2Adaptability or versatility
If the control sequence is made more complicated to support diverse recognition processing, then the adaptability is improved, but the device complexity increases
Solution Approach 1:
The patent extracts the operational sequence complexity from the control unit's hardwired logic and relocates it to external parameter sequences. By separating the control mechanism from the specific operation sequences, the control unit itself remains simple while the diverse recognition processing capabilities are encoded in externally supplied parameter sequences that can be loaded as needed without complicating the control unit's internal structure.
Solution Approach 2:
The patent introduces parameter sequences as an intermediary layer between the control unit and the diverse recognition processing operations. These parameter sequences act as a mediator that translates high-level recognition task requirements into specific control signals, allowing the simple control unit to orchestrate complex recognition operations without directly embedding the complexity within the control unit itself.
3Productivity
If separate hardware is provided for multiply-accumulate operation and post-processing operation, then the productivity is improved, but the device complexity increases
Solution Approach 1:
The patent merges the control mechanisms for multiply-accumulate operations and post-processing operations into a single unified control unit. Both operation types share the same control structure, parameter loading mechanism, and data flow paths, allowing the system to maintain high processing efficiency through specialized hardware units while avoiding the complexity of separate control systems for each operation type.
Solution Approach 2:
The control unit is designed with universal functionality to handle both multiply-accumulate operations and post-processing operations through a single interface. The same control unit loads parameter sequences and directs data flow for different operation types, eliminating the need for separate control hardware while maintaining the productivity benefits of having dedicated processing units for each operation type.
Data Source
AI summary
A data processing apparatus includes a multiply-accumulate operation unit configured to perform multiply-accumulate operation on data based on a given multiply-accumulate operation parameter, an arithmetic operation result retention unit configured to retain an arithmetic operation result obtained by the multiply-accumulate operation unit, a command processing unit configured to, based on a given command, perform processing defined by the command on the arithmetic operation result retained by the arithmetic operation result retention unit, and a distribution unit configured to receive a parameter sequence obtained by a combination of the multiply-accumulate operation parameter and the command and configured to, based on a content and sequential order of the combination, give the multiply-accumulate operation parameter to the multiply-accumulate operation unit and give the command to the command processing unit.