Deterministic Stochastic Computing with RNS for Faster Accurate Arithmetic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional deterministic stochastic computing (SC) methods suffer from long processing time and high energy consumption due to the need to process high-bit-width binary numbers, which limits their application.
Innovation Solution
The integration of deterministic SC methods with the Residue Number System (RNS) reduces the length of bit-streams exponentially, leveraging the parallelism and modularity of RNS to perform arithmetic operations on smaller residues, thereby reducing processing time and energy consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional deterministic stochastic computing methods are used to process high-bit-width binary numbers, then computation accuracy is maintained, but processing time and energy consumption increase significantly
Solution Approach 1:
The patent segments high-bit-width binary numbers into multiple smaller residue components using the Residue Number System. Each residue is processed independently through separate stochastic bit-streams, allowing parallel computation. This segmentation maintains computation accuracy while reducing the processing time exponentially by avoiding the need to process the entire high-bit-width number as a single unit.
Solution Approach 2:
The patent transforms the computation from a single-dimensional high-bit-width binary processing approach into a multi-dimensional parallel processing architecture. By mapping binary numbers to residue components across multiple dimensions (parallel bit-streams), the system achieves faster processing while maintaining accuracy through the mathematical properties of the Residue Number System.
2Measurement precision
If conventional deterministic stochastic computing methods are used to process high-bit-width binary numbers, then computation accuracy is maintained, but energy consumption increases significantly
Solution Approach 1:
The patent segments high-bit-width binary numbers into multiple smaller residue components using the Residue Number System. Each residue is processed independently through separate stochastic bit-streams, allowing parallel computation. This segmentation maintains computation accuracy while reducing the processing time exponentially by avoiding the need to process the entire high-bit-width number as a single unit.
Solution Approach 2:
The patent transforms the computation from a single-dimensional high-bit-width binary processing approach into a multi-dimensional parallel processing architecture. By mapping binary numbers to residue components across multiple dimensions (parallel bit-streams), the system achieves faster processing while maintaining accuracy through the mathematical properties of the Residue Number System.
3Measurement precision
If the length of bit-streams is increased to improve accuracy in stochastic computing, then computation accuracy improves, but latency and energy consumption increase
Solution Approach 1:
The patent segments high-bit-width binary numbers into multiple smaller residue components using the Residue Number System. Each residue is processed independently through separate stochastic bit-streams, allowing parallel computation. This segmentation maintains computation accuracy while reducing the processing time exponentially by avoiding the need to process the entire high-bit-width number as a single unit.
Data Source
AI summary
Inaccuracy of computations is an important challenge with the Stochastic Computing (SC) paradigm. Recently, deterministic approaches to SC are proposed to produce completely accurate results with SC circuits. Instead of random bit-streams, the computations are performed on structured deterministic bit-streams. However, current deterministic methods take a large number of clock cycles to produce correct result. This long processing time directly translates to very high energy consumption. This invention proposes a design methodology based on the Residue Number Systems (RNS) to mitigate the long processing time of the deterministic methods. Compared to the state-of-the-art deterministic methods of SC, the proposed approach delivers improvements in terms of processing time and energy consumption.