Reconfigurable Adder-Multiplier Architecture for AI and Communication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current processing apparatuses are optimized separately for communication algorithms and AI algorithms, which are based on complex numbers and real numbers respectively, leading to inefficiencies in computing, area, and energy usage.
Innovation Solution
A processing apparatus with a computing unit and a control unit that supports both real-number-stream-based and complex-number-stream-based operations through flexible connection relationships between adders and multipliers, allowing for efficient execution of both communication and AI algorithms on a single hardware platform.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If separate processing apparatuses are designed for communication algorithms and AI algorithms, then each algorithm type can be optimized for its specific mathematical model (complex numbers for communication, real numbers for AI), but the device complexity increases and hardware resources are not efficiently utilized
Solution Approach 1:
The processing apparatus is designed with a universal computing unit that can handle both complex-number-based communication algorithms and real-number-based AI algorithms. The computing unit includes configurable computing elements that can be dynamically adjusted to match the mathematical model requirements of different algorithm types, eliminating the need for separate specialized processing apparatuses.
Solution Approach 2:
The processing apparatus employs dynamic configuration of computing elements that can switch between different connection relationships and operational modes. This dynamic adaptability allows the same hardware to efficiently process both communication algorithms (requiring complex number operations) and AI algorithms (requiring real number operations) without requiring separate fixed-architecture processors.
2Reliability
If separate processing apparatuses are used for communication and AI algorithms, then each can be optimized for its specific operations, but the area efficiency decreases due to duplicate hardware resources
Solution Approach 1:
The patent merges the functionality of separate communication processing apparatus and AI processing apparatus into a single unified processing apparatus. The computing unit integrates both complex-number operation capabilities and real-number operation capabilities, allowing both algorithm types to share the same hardware resources including computing elements, memory, and control logic, thereby reducing overall hardware area while maintaining algorithm-specific optimization.
3Productivity
If separate processing apparatuses are deployed for communication and AI algorithms, then each can operate at optimal efficiency for its specific model, but the energy efficiency decreases due to redundant hardware operation
Solution Approach 1:
The unified processing apparatus provides universal computing capabilities that can adapt to both communication algorithm requirements (complex number operations) and AI algorithm requirements (real number operations). This eliminates the need for two separate processing apparatuses operating in parallel, thereby reducing redundant energy consumption while maintaining optimal computing efficiency for both algorithm types through dynamic configuration of the computing elements.
Data Source
AI summary
A processing apparatus and a control method are provided including a computing unit and a control unit. The computing unit includes a processing element, and the processing element supports a first connection relationship and a second connection relationship between at least one adder and at least one multiplier. The first connection relationship is used to implement a real-number-stream-based communication algorithm or artificial intelligence algorithm, and the second connection relationship is used to implement a complex-number-stream-based communication algorithm or artificial intelligence algorithm. The control unit is configured to control a connection relationship used by the processing element, and the connection relationship includes the first connection relationship and the second connection relationship.