Block Floating-Point Conversion Without Recompression Data Loss
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data compression and decompression processes in block floating point (BFP) systems result in data loss and inaccuracies, complicating the conversion between different BFP formats.
Innovation Solution
A method and system for converting BFP data blocks from one format to another without decompressing and compressing, allowing for efficient conversion by determining bit strings and exponents based on output size dependencies, and performing actions such as appending bit strings, shifting, and padding to maintain data integrity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data compression and decompression processes are used in BFP systems, then data transmission efficiency is improved, but data loss and inaccuracies occur
Solution Approach 1:
The patent extracts the essential BFP format conversion requirement from the complex compression-decompression process. By identifying that only format conversion is needed rather than full compression-decompression cycles, the system removes unnecessary steps that cause data loss while maintaining transmission efficiency.
Solution Approach 2:
Instead of following the conventional compression-then-decompression approach, the patent inverts the process by performing direct format conversion between BFP representations. This reversal eliminates the information loss inherent in compression-decompression cycles while achieving the same goal of enabling data transmission.
2Quantity of substance
If conventional BFP compression and decompression algorithms are used, then data sample reduction is achieved, but system complexity increases
Solution Approach 1:
The patent extracts only the necessary format conversion operation from the full compression-decompression algorithm suite. By taking out just the essential conversion logic and eliminating redundant compression/decompression steps, the system reduces data samples effectively while minimizing computational complexity.
Solution Approach 2:
Rather than implementing complete compression and decompression algorithms, the patent applies partial action by performing only the necessary format conversion. This selective approach achieves sufficient data sample reduction without the excessive complexity of full compression-decompression cycles.
3Loss of time
If BFP format conversion is performed without decompression, then processing time is reduced, but conversion accuracy may be compromised
Solution Approach 1:
The patent performs preliminary analysis of the BFP format structures to identify direct conversion pathways. By pre-determining the conversion rules and bit manipulation operations needed, the system achieves rapid format conversion without decompression while maintaining accuracy through carefully designed conversion algorithms.
Solution Approach 2:
The patent changes the representation parameters directly between BFP formats through controlled bit manipulation and exponent adjustment. By transforming the data parameters directly without converting to decompressed form, the system reduces processing time while preserving conversion accuracy through mathematically sound parameter transformations.
Data Source
AI summary
A method, system and apparatus are disclosed. A computing device is described. The computing device comprises processing circuitry configured to convert a first data block comprising a first plurality of data samples in a first format to a second data block comprising a second plurality of data samples in a second format. The conversion is performed without at least one of decompressing and compressing the first plurality of data samples. Further, the conversion is one of output size dependent and output size independent. In addition, the processing circuitry is configured to perform at least one action based on the conversion of the first data block.