MAC Circuit Windowing for Aligned Low-Power DSP Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multiply-accumulate (MAC) circuit units in digital signal processors face issues with high power consumption, misaligned memory accessing, and complex structure, which hinder efficient data processing.
Innovation Solution
The method involves retrieving datasets from memory registers, storing them in specific register portions, generating a replica of coefficients, and performing MAC operations using subsets of dataset elements within windows of equal size, with a shift-setting parameter determining the selection of elements and alignment, thereby optimizing memory access and reducing power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If existing MAC circuit units are used for data processing, then multiplication and accumulation operations can be performed, but power consumption is high
Solution Approach 1:
The patent divides the MAC operation into multiple stages: data retrieval from memory, storage in register portions, generation of coefficient replicas, and performance of MAC operations on subset elements. This segmentation allows for optimized power management at each stage while maintaining computational throughput.
Solution Approach 2:
The patent changes operational parameters by using a shift-setting parameter to control window sizes and element selection. By dynamically adjusting these parameters, the circuit can operate in different modes that balance power consumption and computational efficiency based on specific processing requirements.
2Ease of operation
If existing MAC circuit units are used for data processing, then computational operations can be performed, but memory accessing is misaligned
Solution Approach 1:
The patent performs preliminary actions by retrieving multiple datasets from memory registers and storing them in register portions before the actual MAC operations. This pre-positioning of data ensures proper alignment is achieved beforehand, eliminating the need for complex alignment logic during the critical MAC computation phase.
Solution Approach 2:
The patent introduces register portions as intermediary storage between memory and the MAC computation units. These intermediaries buffer the data and coefficients, allowing the MAC unit to operate on properly aligned data without requiring the memory system itself to be complex or reconfigurable.
3Device complexity
If existing MAC circuit units are used for data processing, then multiply-accumulate operations can be performed, but structure and logic are complex
Solution Approach 1:
The patent creates a universal MAC circuit structure that can handle multiple datasets and coefficient sets through the use of windowing and subset selection mechanisms. The same basic MAC unit can process different data combinations by adjusting the window parameters and element selection, eliminating the need for multiple dedicated circuits.
Solution Approach 2:
The patent introduces dynamic control through the shift-setting parameter that adjusts window sizes and element selection during operation. This dynamic behavior allows a single static circuit structure to adapt to different computational requirements, reducing overall system complexity while maintaining high computational throughput.
Data Source
AI summary
A method includes retrieving a plurality of datasets from respective memory registers of a memory and storing the retrieved plurality of datasets in respective register portions of a first register. A dataset of data-processing coefficients are stored in a second register. First processing is applied using, as the first operand, a first sub-set of dataset elements stored in the first register, and using, as the second operand, the data-processing coefficients, obtaining a first result. Second processing is applied using, as the first operand, a second sub-set of dataset elements stored in the first register comprised in a second window having a size equal to the dataset size, and using, as the second operand, the replica of the dataset of data-processing coefficients, obtaining a second result. An output is generated based on the first and second results. The first and second processing may perform multiply accumulate (MAC) operations.