Discrete Store Address Mapping for Lower-Overhead Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face complexity and high data overhead in processing discrete data points during image recognition, leading to inefficient processing processes.
Innovation Solution
A data processing method that determines source and discrete destination addresses for continuous data, allowing sequential storage of data to obtain discrete data points, simplifying the processing and reducing overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If discrete data points are processed by means of related technologies, then data processing can be performed, but the processing process becomes complicated and data overhead increases
Solution Approach 1:
The patent segments the processing of discrete data points by introducing a vectorization mechanism. Instead of processing each discrete data point individually through complicated routines, the system segments data into vectors that can be processed in bulk, thereby simplifying the overall processing process while maintaining productivity.
Solution Approach 2:
The patent transitions from processing data in discrete scalar form to processing data in vector form, effectively adding a dimensional aspect to data representation. This dimensionality change allows multiple discrete data points to be handled simultaneously as a unified vector structure, reducing processing complexity and overhead.
2Productivity
If discrete data points are processed by means of related technologies, then data processing can be performed, but data overhead becomes large
Solution Approach 1:
The patent merges multiple discrete data points into a unified vector structure. By combining individual data points into vectors, the system reduces the overhead associated with handling each discrete point separately, while still maintaining the capability to process all necessary data points efficiently.
Solution Approach 2:
By transitioning from scalar to vector representation, the patent reduces data overhead through more compact and efficient data structures. The vector format allows for reduced redundancy and more efficient memory utilization compared to processing discrete data points individually.
3Ease of operation
If vectors are restored to discrete data points after vector computation processing, then discrete data can be obtained, but the processing process becomes complicated
Solution Approach 1:
The patent applies preliminary action by maintaining vector structures throughout the computation process and only converting to discrete data points when absolutely necessary for output or specific operations. This preliminary conversion approach avoids repeated and complicated vector-to-discrete transformations during intermediate processing steps.
Solution Approach 2:
The vector structure serves as an intermediary between discrete data input and discrete data output. By using vectors as the intermediate representation throughout the computation pipeline, the system avoids the complexity of repeated conversions, with the vector acting as a mediator that simplifies the overall data flow and processing operations.
Data Source
AI summary
The present disclosure relates to a data processing method, a data processing apparatus, and related products. The data processing apparatus includes an address determining unit and a data storage unit. The address determining unit is configured to determine a source data address and a plurality of discrete destination data addresses of data corresponding to a processing instruction when the decoded processing instruction is a discrete store instruction, where the source data address may include continuous data addresses. The data storage unit is configured to obtain a plurality of pieces of discrete data by sequentially storing data read from the source data address to the plurality of destination data addresses.