Entropy Coding with Fixed State Thresholds for SIMD Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current entropy coding methods, such as Asymmetric Numeral System (ANS) and Arithmetic Coding (AC), are computationally and memory-intensive due to their iterative approaches, which hinder performance in single instruction, multiple data (SIMD) parallel implementations and increase latency in data compression and decompression processes.
Innovation Solution
Implementing a noniterative entropy encoding and decoding technique that uses fixed state thresholds to manage the state variable within bounds, eliminating the need for conditional loops and reducing computational and memory requirements by determining the number of bits to remove or add based on these thresholds during encoding and decoding processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If iterative entropy coding methods (ANS/AC) are used to achieve optimal compression ratios, then compression efficiency is improved, but computational complexity and memory bandwidth requirements increase
Solution Approach 1:
The state variable is segmented into fixed-width fields (e.g., 16-bit or 32-bit segments) that can be processed independently. This segmentation allows the encoding operation to be divided into discrete, parallelizable steps that manipulate specific bit segments rather than requiring full iterative processing of the entire state.
Solution Approach 2:
The iterative arithmetic coding mechanism is replaced with a direct bit-manipulation approach using bitwise operations. Instead of repeatedly applying arithmetic operations to converge on the encoded value, the patent uses single-step bitwise shifts, masks, and logical operations to achieve the same entropy coding effect without iteration.
2Loss of information
If iterative entropy coding methods are used to achieve optimal compression ratios, then compression efficiency is improved, but processing speed decreases
Solution Approach 1:
The necessary bit manipulations are prepared and executed in a predetermined sequence using fixed-width state variables. The encoding process performs all necessary operations in advance within a known number of steps, eliminating the need for iterative convergence and enabling predictable, high-speed processing.
Solution Approach 2:
Iterative arithmetic operations are replaced with direct bitwise operations that execute in constant time. The patent uses bitwise shifts, masks, and logical operations to perform entropy coding in a single pass rather than requiring multiple iterative steps, dramatically increasing processing speed.
3Loss of information
If iterative entropy coding methods are used to achieve optimal compression ratios, then compression efficiency is improved, but parallelization capability is reduced
Solution Approach 1:
The state variable is divided into independent fixed-width segments that can be processed in parallel. Each segment can be encoded independently using the same bitwise operations, allowing multiple segments to be processed simultaneously using SIMD (Single Instruction Multiple Data) instructions or multi-threaded architectures.
Solution Approach 2:
The fixed-width state variable approach creates a universal encoding mechanism that can be applied identically to multiple data elements simultaneously. The same bitwise operations work on any fixed-width state, enabling the encoding function to be replicated and executed in parallel across multiple data streams or image blocks.
4Loss of information
If iterative entropy coding methods are used to achieve optimal compression ratios, then compression efficiency is improved, but latency increases
Solution Approach 1:
All necessary encoding operations are performed in a predetermined sequence within a fixed number of steps. The patent prepares and executes the complete encoding process in advance using fixed-width state variables, eliminating the variable-time iterative convergence and enabling constant-latency processing suitable for real-time applications.
Solution Approach 2:
The time-consuming iterative arithmetic operations are replaced with fast bitwise operations that execute in constant time. The patent uses single-cycle bitwise shifts, masks, and logical operations to achieve entropy coding without the multiple iterative steps required by traditional methods, dramatically reducing processing latency.
Data Source
AI summary
This disclosure provides methods, devices, and systems for data compression and decompression. The present implementations more specifically relate to entropy encoding and decoding techniques for keeping a state variable within upper and lower bounds using a noniterative process. The entropy encoding uses a fixed state threshold to determine a number of bits to remove and removes the bits from a current state prior to encoding a symbol with the current state. The entropy decoding decodes encoded data in a bitstream based on a current state to obtain the symbol and a new state and determines a number of bits to read from the bitstream and to add to the new state to update the current state.