Multi-Processor Data Packing Using Sensitivity-Based Bit Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computer systems face challenges in reducing bit length for numerical data representation, leading to increased storage and transmission costs, power consumption, and processing time, particularly in distributed computing systems where data movement across memory hierarchies is costly in terms of time and energy.
Innovation Solution
The implementation of sensitivity analysis and bit elimination techniques, where the sensitivity parameter identifies and eliminates 'garbage' bits in binary representations of numbers, allowing for dynamic adaptation of bit length based on the sensitivity of data elements during arithmetic operations and data movement, thereby reducing the number of bits required for storage and transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full precision bit length is used for numerical data representation, then computational accuracy is maintained, but storage and transmission costs increase
Solution Approach 1:
The patent applies local quality by differentiating the precision requirements of different data elements within the same computational system. Instead of uniformly using full precision for all numerical data, the system dynamically determines the actual precision needed for each specific data element based on its sensitivity characteristics. This allows less precise representation for data elements that don't require full precision, thereby reducing storage and transmission costs while maintaining computational accuracy where necessary.
Solution Approach 2:
The patent implements parameter changes by dynamically adjusting the bit length parameter for numerical data representation. The system monitors sensitivity parameters during computation and adapts the precision (bit length) of data elements in real-time. When sensitivity analysis shows that fewer bits are sufficient to maintain computational accuracy, the system reduces the bit length parameter, thereby reducing storage and transmission requirements without compromising the required computational precision.
2Productivity
If data movement across memory hierarchies is increased, then processing capability is enhanced, but time and energy consumption increase
Solution Approach 1:
The patent applies the extraction principle by removing unnecessary data elements from computation. Through sensitivity analysis, the system identifies data elements with zero or negligible sensitivity that do not contribute to the computational result. By extracting and eliminating these redundant data elements before they enter the computation pipeline, the system reduces the volume of data that needs to be moved across memory hierarchies, thereby reducing time and energy consumption while maintaining processing capability for the essential data.
3Quantity of substance
If bit length is reduced for data representation, then storage and transmission efficiency improve, but computational accuracy may deteriorate
Solution Approach 1:
The patent implements feedback through sensitivity analysis that continuously monitors the impact of bit length reduction on computational accuracy. The system calculates sensitivity parameters that indicate how much each data element contributes to the final result. This feedback mechanism allows the system to adjust bit length decisions dynamically - if reducing bit length for a data element would significantly impact computational accuracy (high sensitivity), the system maintains full precision; if the impact is negligible (low sensitivity), the system safely reduces bit length to improve storage and transmission efficiency.
Data Source
AI summary
Data employed in computations amongst multiple processors in a computing system is processed so that less bits than a full representation of the data needs to be communicated yet all of the relevant information is communicated. More specifically, a sensitivity value is determined for various items of the data which reflect the number of bits in the data items that are not garbage bits, and only information in the data items that are indicated by the sensitivity value to not be garbage bits are necessarily effectively retained. At least the information that is not garbage bits and the corresponding associated sensitivity are packed together. The results of computations that are performed using the data items as at least one of the operands for the computation are associated with a sensitivity that is derived from the individual sensitivities of the operands used in the computation.