Video Decoder Parallel Processing via Segmented Buffer Memory
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing moving-picture decoding apparatuses for H.265/HEVC and H.264/AVC standards face a reduction in parallel processing capability due to the need for a no-operation instruction, which leads to overwriting of reference picture information and inefficient pipeline operations, especially when handling large pixel sizes and high-definition video signals.
Innovation Solution
The proposed solution involves a moving-picture decoding processing apparatus with a decoding control unit and multiple processing units that manage frame decoding and processing stages differently, inhibiting the use of certain signals at specific periods to prevent reference picture overwriting and optimize pipeline operations, allowing for improved parallel processing by executing decoding processes without no-operation instructions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If parallel processing is implemented in existing moving-picture decoding apparatuses, then processing speed is improved, but reference picture information is overwritten causing processing errors
Solution Approach 1:
The decoding apparatus is divided into multiple independent decoding processing units (first decoding processing unit 20, second decoding processing unit 21) that operate in parallel. Each unit has dedicated buffer memory (240, 241) to store reference picture information separately, preventing overwriting while maintaining parallel processing capability.
Solution Approach 2:
Buffer memory units (240, 241) are introduced as intermediaries between the decoding processing units and the frame memory. These buffer memory units temporarily store reference picture information, allowing parallel decoding operations to proceed without causing data overwriting or conflicts.
2Reliability
If no-operation instructions are inserted to prevent overwriting, then reference picture integrity is maintained, but pipeline operations become inefficient
Solution Approach 1:
The buffer memory is segmented into multiple independent units (240, 241) that can be accessed simultaneously by different decoding processing units. This segmentation eliminates the need for no-operation instructions because each unit has dedicated storage space, allowing continuous pipeline operations without interruptions.
Solution Approach 2:
Reference picture information is preliminarily stored in the buffer memory units before being needed by subsequent decoding operations. This preliminary action ensures that data is ready and available when needed, maintaining continuous pipeline flow without requiring停顿 for data preparation or protection.
3Productivity
If multiple decoding processing units operate in parallel, then decoding throughput is improved, but resource conflicts occur
Solution Approach 1:
Resources are segmented into dedicated units for each decoding processing unit. Buffer memory units (240, 241) are specifically assigned to specific decoding processing units (20, 21), creating independent resource domains that eliminate conflicts while enabling parallel operation. This segmentation simplifies resource management compared to shared resources requiring complex arbitration.
Solution Approach 2:
Each buffer memory unit serves multiple functions: storing reference picture information for the associated decoding processing unit, providing data for motion compensation operations, and supporting both intra-picture and inter-picture decoding. This multi-functionality reduces the need for separate specialized resources for each function.
Data Source
AI summary
The present invention is directed to reduce deterioration in parallel processing capability. In a moving-picture decoding processing apparatus, information of a first plurality of frames and a second plurality of frames is supplied from a decoding control unit to first and second decoding processing units. For decoding the information from an intermediate point to an end point of a second preceding frame by the second decoding unit in a third period, use of a result of the process of the first decoding processing unit in the third period is inhibited, and use of a result of the process of a first preceding frame by the first decoding processing unit in a second period is permitted by an end signal. By reach of the intermediate point of the decoding of the second preceding frame by the second decoding processing unit in the second period, information from the start point to the intermediate point of the first subsequent frame included in the plural frames is decoded by the first decoding unit in the third period.