Blockwise Data Compression for Precision-Aware Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing processing systems face inefficiencies in computational complexity, power consumption, and processing time due to the varying bit sizes and types of numerical data used in operations, particularly in machine learning tasks, leading to reduced battery life and processing delays.
Innovation Solution
Implementing hierarchical blockwise vector data compression techniques that identify and remove leading bits and truncate least significant bits, using an arithmetic shifter and multiplexer to add a carry bit, allowing for efficient data compression and decompression in processing systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If larger bit size data (e.g., 64-bit integers, 32-bit floating-point numbers) is used for mathematical operations, then computational precision and accuracy are improved, but computational complexity increases
Solution Approach 1:
The patent dynamically changes the bit size parameter of data based on the specific computational task requirements. The system selects appropriate data precision levels (e.g., 8-bit, 16-bit, 32-bit, or 64-bit) for different operations, allowing computational complexity to be adjusted while maintaining sufficient precision for each specific task
Solution Approach 2:
The patent segments mathematical operations into different precision levels based on their requirements. Different data types and operations are categorized into groups that require specific bit sizes, allowing the system to apply appropriate precision only where needed rather than uniformly across all operations
2Measurement precision
If more computationally complex operations are executed, then processing accuracy is improved, but power consumption increases
Solution Approach 1:
The system dynamically adjusts data bit size parameters based on operational requirements, using lower precision (smaller bit size) when high accuracy is not needed and higher precision only when necessary. This parameter adaptation reduces power consumption by avoiding unnecessarily complex operations while maintaining required processing accuracy
3Measurement precision
If more computationally complex operations are executed, then processing accuracy is improved, but processing time increases
Solution Approach 1:
The patent implements dynamic parameter adjustment by changing data bit size based on the specific computational task. By selecting the minimum necessary precision level for each operation, the system reduces processing time while maintaining sufficient accuracy, avoiding the overhead of processing high-precision data when it is not required
4Quantity of substance
If data is compressed by removing leading bits and truncating least significant bits, then data size is reduced, but data precision is lost
Solution Approach 1:
The patent applies different compression strategies to different parts of the data based on their importance. Leading bits that contribute most to precision are preserved, while least significant bits that have minimal impact on accuracy are truncated. This local differentiation allows data size reduction while maintaining essential precision
Solution Approach 2:
The system dynamically adjusts the compression level by varying the number of bits to remove and truncate based on the specific data characteristics and computational requirements. This parameter adaptation allows optimal balance between data size reduction and precision preservation for different data types and operations
Data Source
AI summary
Certain aspects of the present disclosure provide techniques and apparatus for efficiently performing operations using data compression. An example method generally includes identifying, for a block of data samples, a number of leading bits to remove from each data sample in the block of data samples. A block of compressed data samples is generated based on the identified number of leading bits and truncation of a number of least significant bits from each data sample in the block of data samples. A bitstream including the block of compressed data samples and an indication of a type of compression applied to the block of data samples is generated and output for further processing.