2D Memory Word Layout for Low-Bandwidth Video Prediction Access
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-definition video applications face challenges in memory bandwidth constraints, as conventional memory word organizations require excessive memory access for retrieving predictions, leading to inefficient use of memory bandwidth, especially when prediction sizes exceed memory bus widths.
Innovation Solution
Implementing a 2D subarray memory word organization where each memory word corresponds to a 2D subarray, allowing for the retrieval of multiple predictions as part of a larger array, reducing the average and worst-case bandwidth required, and combining multiple predictions in a single memory access operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional memory word organization is used, then memory access is simple, but memory bandwidth efficiency deteriorates due to excessive memory access for retrieving predictions
Solution Approach 1:
The patent segments the memory word into multiple 2D subarrays, where each subarray can independently store prediction data. This segmentation allows the memory system to retrieve multiple predictions simultaneously by accessing different subarrays within a single memory word, thereby improving memory bandwidth efficiency without requiring complex external memory access operations
Solution Approach 2:
The patent transitions from conventional one-dimensional linear memory organization to two-dimensional subarray organization within memory words. This dimensional change enables spatial mapping of prediction data that matches the 2D nature of video blocks, allowing efficient retrieval of multiple predictions by exploiting the 2D structure and reducing the number of memory access operations required
2Productivity
If multiple predictions are retrieved individually, then memory access control is simple, but memory bandwidth efficiency deteriorates due to multiple separate access operations
Solution Approach 1:
The patent merges multiple prediction data structures into a single memory word by organizing them as 2D subarrays. This merging allows the memory system to retrieve multiple predictions that would traditionally require separate memory access operations, now contained within a single memory word access, thereby improving prediction retrieval efficiency while maintaining simple memory access control
Solution Approach 2:
The memory word organization provides multi-functionality by enabling a single memory word to store and deliver multiple different prediction data structures simultaneously. This universal structure can accommodate various prediction types and sizes, allowing the memory system to serve multiple prediction retrieval needs through a single standardized access operation
3Quantity of substance
If memory bus width is increased, then data transfer capacity improves, but memory bandwidth efficiency deteriorates when prediction sizes exceed bus widths
Solution Approach 1:
The patent segments prediction data into 2D subarrays that fit within memory word structures, allowing efficient utilization of available memory bus width. By organizing prediction data to match the 2D subarray structure, the system maximizes the data transfer capacity of the memory bus without leaving unused bandwidth, even when prediction sizes vary
Data Source
AI summary
Described systems and methods allow a reduction in the memory bandwidth required in video coding (decoding/encoding) applications. According to a first aspect, the data assigned to each memory word is chosen to correspond to a 2D subarray of a larger array such as a macroblock. An array memory word organization allows reducing both the average and worst-case bandwidth required to retrieve predictions from memory in video coding applications, particularly for memory word sizes (memory bus widths) larger than the size of typical predictions. According to a second aspect, two or more 2D subarrays such as video predictions are retrieved from memory simultaneously as part of a larger 2D array, if retrieving the larger array requires fewer clock cycles than retrieving the subarrays individually. Allowing the combination of multiple predictions in one memory access operation can lead to a reduction in the average bandwidth required to retrieve predictions from memory.