Compressed Polar ECC Decoding for Low Gate-Count Non-Dyadic Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional compression techniques for non-dyadic values in computer systems result in high gate-count compressors and decompressors, and variable-length codewords that do not guarantee gate count reduction, leading to power consumption and routing issues due to increased complexity.
Innovation Solution
A storage system that groups encoded values into codewords with variable-length prefixes and bit chunks of predetermined lengths, allowing for efficient compression and decompression using a fractional compression method that reduces gate-count overhead and supports fixed-to-fixed compression, enabling memory and bus savings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of substance
If conventional compression techniques (Huffman coding, arithmetic coding) are used for non-dyadic values, then data compression is achieved, but gate-count complexity increases significantly
Solution Approach 1:
The patent segments the compression task into two distinct phases: a compression-free phase that processes a prefix of the input sequence without any compression operations, and a compression phase that applies compression techniques only to the remaining suffix. This segmentation avoids the high gate-count complexity of conventional compression by isolating it to a smaller portion of the data, thereby reducing overall device complexity while still achieving redundancy reduction.
Solution Approach 2:
The patent performs preliminary action by processing a prefix of the input sequence without compression, preparing the data in a form that enables more efficient compression in the subsequent phase. This preliminary uncompressed processing simplifies the initial data structure and reduces the complexity burden before the compression phase begins, effectively lowering the overall gate-count requirements.
2Loss of substance
If conventional compression techniques are used, then data compression is achieved, but power consumption increases due to high gate-count
Solution Approach 1:
By segmenting the compression process into uncompressed prefix and compressed suffix phases, the patent reduces the number of active compression gates operating simultaneously, thereby lowering dynamic power consumption. The segmentation ensures that high-power compression operations are confined to a smaller data portion.
Solution Approach 2:
The patent applies compression partially rather than fully to the entire input sequence. By compressing only the suffix portion while leaving the prefix uncompressed, the system achieves sufficient redundancy reduction without activating the full power-consuming compression circuitry for all data, thus reducing overall power consumption.
3Adaptability or versatility
If variable-length codewords are used to support worst-case data distribution, then data representation flexibility is improved, but gate-count reduction cannot be guaranteed
Solution Approach 1:
The patent segments the input sequence into prefix and suffix portions with different processing strategies. The prefix handles worst-case scenarios without compression, while the suffix applies compression for typical cases. This segmentation allows the system to adapt to different data distributions without requiring complex variable-length codeword circuits for the entire data stream.
Solution Approach 2:
The patent extracts the problematic worst-case scenario handling from the main compression pipeline by dedicating the prefix portion to uncompressed processing. This extraction isolates the adaptability requirements to a small prefix segment, allowing the majority of the system to use simpler fixed-length or reduced-variable-length codewords, thereby reducing overall gate-count complexity.
4Loss of substance
If compression is applied to reduce data width, then memory and bus savings are achieved, but routing complexity increases due to variable-length codewords
Solution Approach 1:
By segmenting the data into prefix and suffix with different compression treatments, the patent creates more predictable data width characteristics. The uncompressed prefix maintains consistent width, while the compressed suffix has reduced but still manageable variable length, overall reducing routing complexity compared to fully variable-length compression.
Solution Approach 2:
The preliminary uncompressed processing of the prefix phase prepares data with more regular width characteristics before the compression phase. This preliminary action reduces the variability introduced by full compression, making subsequent routing and bus operations simpler and more predictable.
Data Source
AI summary
A storage system, including a storage device configured to store a plurality of encoded values, wherein each value of the plurality of encoded values has a predetermined value length and is within a predetermined range, and wherein the predetermined range is not a power of 2; and at least one processor configured to: group the plurality of encoded values into a codeword; obtain a plurality of bit chunks, wherein each bit chunk of the plurality of bit chunks represents a corresponding encoded value of the plurality of encoded values, and wherein a length of the each bit chunk is selected from among one or more predetermined bit chunk lengths which are determined based on the predetermined range; select a variable-length prefix from among a plurality of variable-length prefixes, wherein the variable-length prefix indicates bit chunk lengths of the plurality of bit chunks; obtain a compressed codeword including the variable-length prefix and the plurality of bit chunks; and decode the plurality of encoded values based on the compressed codeword.