Earth Model Data Compression via Look-Up Table Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional compression methods for earth model data in seismic processing operations are inefficient, prioritizing storage reduction over accuracy and computational efficiency, leading to suboptimal performance in data access and processing.
Innovation Solution
A method that compresses earth model datasets using a look-up table and data indices, allowing for selective decompression at a processor, optimizing storage and computational efficiency by storing the look-up table in fast memory and indices in larger, slower memory, enabling rapid access and repeated decompression without external memory access.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If conventional compression methods are used to reduce storage requirements, then data storage capacity is improved, but computational efficiency and data access accuracy deteriorate
Solution Approach 1:
The compressed earth model data is segmented into two distinct components: a look-up table containing quantized data values stored in fast first level memory, and data indices stored in larger second level memory. This segmentation allows the system to leverage the speed advantage of fast memory for frequent look-ups while using capacity advantage of slow memory for storing the complete index structure, thereby resolving the contradiction between storage capacity and computational efficiency.
Solution Approach 2:
Different quality levels of data representation are assigned to different memory locations based on access patterns. The look-up table with quantized values is placed in fast memory because it is frequently accessed during computation, while the full precision indices are stored in slower but larger capacity memory. This local quality differentiation optimizes the balance between speed and storage capacity for different data components.
2Speed
If earth model data is stored in local memory to improve access speed, then data access rate is improved, but the size of data that can be stored is limited
Solution Approach 1:
The system segments the earth model data into a compact look-up table and sparse data indices. The look-up table, which contains frequently accessed quantized values, is stored in fast local memory. The data indices, which are fewer in number and can be stored in larger memory, are placed in second level memory. This segmentation enables the system to store more total data while maintaining fast access rates for the critical look-up operations.
3Quantity of substance
If conventional compression methods are used, then storage requirements are reduced, but decompression cost and complexity increase
Solution Approach 1:
The compression process performs preliminary quantization of earth model data values into a look-up table during an initial compression step. This pre-computed look-up table can then be efficiently accessed during subsequent decompression and computation operations. By performing the complex quantization operation in advance during compression, the system simplifies the decompression process to simple table look-ups, thereby reducing decompression complexity while maintaining storage efficiency.
Data Source
AI summary
A method and corresponding system is provided for computation utilizing an earth model representation via a computing system having a first processor having access to an earth model dataset. The method includes compressing the earth model dataset at the first processor to generate a look-up table and a set of data indices (i.e., collectively a compressed earth model representation), wherein the look-up table includes quantized data values. By then storing the look-up table in a first level (“fast”) memory, and storing the indices in a second level (“slower,” higher memory capacity) memory, the look-up table and the indices can be accessed to selectively decompress the compressed earth model representation at the first processor such that the computation can be performed efficiently by the first processor.