Balanced Throughput Data-Path Architecture for FIR and Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital signal processing architectures face inefficiencies due to bubble cycles caused by memory misalignment during FIR and correlation operations, which reduce overall throughput and prevent multiple MAC operations per cycle.
Innovation Solution
A hierarchical Grouped Register File (GRF) system with a misaligned address placement mechanism and a dedicated data path from the Address Generator to the MAC execution unit, allowing for simultaneous loading of data into grouped registers and bypassing of misaligned data directly into the MAC unit, enabling balanced throughput and eliminating bubble cycles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a balanced throughput architecture is used for FIR operations, then data throughput is balanced, but bubble cycles occur due to memory misalignment during correlation operations
Solution Approach 1:
The register file is segmented into multiple ports (first port and second port) with different functions. The first port handles balanced throughput operations while the second port handles misaligned address operations. This segmentation allows the system to handle different memory access patterns simultaneously without interference, eliminating bubble cycles while maintaining balanced throughput.
Solution Approach 2:
The address generator dynamically switches between two modes of operation: balanced throughput mode for normal FIR operations and misaligned address mode for correlation operations. This dynamic adaptation allows the architecture to optimize data loading based on the specific operation requirements, preventing bubble cycles during misaligned access while maintaining overall throughput efficiency.
2Productivity
If data is loaded into register file for MAC operations, then computations can proceed, but misaligned addresses cause delays and reduce throughput
Solution Approach 1:
A second port is introduced as an intermediary mechanism to handle misaligned address operations. When misaligned addresses occur, the second port loads the required data directly without requiring the standard balanced throughput path, acting as a mediator that resolves access delays caused by memory misalignment while maintaining overall system throughput.
3Productivity
If multiple MAC operations are performed per cycle, then productivity increases, but memory access balance becomes difficult to maintain
Solution Approach 1:
The data path is segmented into multiple independent ports that can operate simultaneously. The first port handles balanced throughput operations while the second port handles misaligned operations, allowing multiple MAC operations to be performed per cycle without increasing overall path complexity. Each port maintains its own simplicity while contributing to the overall productivity enhancement.
Data Source
AI summary
Apparatus and method for a modified, balanced throughput data-path architecture is given for efficiently implementing the digital signal processing algorithms of filtering, convolution and correlation in computer hardware, in which both data and coefficient buffers can be implemented as sliding windows. This architecture uses a multiplexer and a data path branch from the Address Generator unit to the multiply-accumulate execution unit. By selecting between the data path of Address Generator to execution unit and the data path of register to execution unit, the unbalanced throughput and multiply-accumulate bubble cycles caused by misaligned addressing on coefficients can be overcome. The modified balanced throughput data-path architecture can achieve a high multiply-accumulate operation rate per cycle in implementing digital signal processing algorithms.