Processor-Core Search Unit for Faster Variable-Length Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern processors face inefficiencies in processing complex instructions, particularly in compression and decompression operations, which hinder overall processor performance and throughput due to the time-consuming nature of graph traversal and inter-node communications in large systems.
Innovation Solution
Implementing a search unit within the processor core that utilizes a look-up table and a finite state machine to accelerate compression and decompression processes by identifying appropriate compression or decompression symbols, with the table sorted by uncompressed or compressed symbols to facilitate faster searches and traversal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If graph traversal is used for compression and decompression operations, then the operations can be performed, but the execution time becomes excessively long due to the complexity of identifying compression symbols
Solution Approach 1:
The patent pre-sorts the compression symbol table by symbol value before compression operations begin. This preliminary arrangement of data allows for direct binary search during compression, eliminating the need for complex graph traversal at runtime and dramatically reducing execution time while maintaining correct compression operations
Solution Approach 2:
The patent replaces the mechanical graph traversal algorithm with a binary search algorithm operating on a pre-sorted table. This substitution changes the fundamental approach from sequential graph-based symbol identification to efficient logarithmic-time binary search, resolving the time complexity issue
2Productivity
If graph traversal is used for symbol identification, then compression can be performed, but processor throughput is reduced due to the time-consuming nature of the operation
Solution Approach 1:
The compression symbol table is pre-sorted by symbol value before use. This preliminary data preparation enables binary search during compression operations, reducing the time each compression instruction takes to execute and thereby increasing overall processor throughput for compression workloads
Solution Approach 2:
The patent changes the search parameter from linear graph traversal to logarithmic binary search on a sorted table. This parameter change in the search algorithm's time complexity directly reduces instruction execution duration and improves processor throughput for compression and decompression operations
3Loss of time
If a sorted look-up table is used to accelerate compression, then execution time is reduced, but the device complexity increases due to the need for additional data structures
Solution Approach 1:
The patent creates a sorted copy of the compression symbol table for use during compression operations. This copied and sorted data structure enables fast binary search while keeping the original symbol definition table intact, resolving the time-speed tradeoff with minimal impact on device complexity
Solution Approach 2:
The patent adds a sorting dimension to the compression symbol table, organizing symbols by their value rather than by their definition order. This additional organizational dimension enables efficient binary search without fundamentally changing the core compression algorithm, achieving speed improvement with acceptable complexity increase
Data Source
AI summary
Systems and methods to accelerate compression and decompression with a search unit implemented in the processor core. According to an embodiment, a search unit may be implemented to perform compression or decompression on an input stream of data. The search unit may use a look-up table to identify appropriate compression or decompression symbols. The look-up table may be populated with a table derived using the variable length coding symbols of a sequence of vertices to be compressed or extracted from a received data stream to be decompressed. A comparator and a finite state machine may be implemented in the search unit to facilitate traversal of the look-up table.