Video Decoder Prefetch Buffer for Intra Block Copy Latency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video decoding technologies face performance bottlenecks due to memory latency issues during intra block copy operations, which can stall the pipeline and slow down video decoding processes, especially when accessing slower memory storage units.

Innovation Solution

The implementation of a video decoder architecture that combines a local buffer with a temporal prediction unit, an intra block control unit, and a prefetch buffer to manage memory requests, allowing for prefetching of source data and reducing memory latency by storing it in a faster local buffer, thereby maintaining pipeline performance and enabling real-time video playback.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If data is stored in slower memory storage units, then memory capacity is improved, but memory access speed deteriorates causing pipeline stalls

Engineering Contradiction:
Improvememory capacityVSAvoidmemory access speed
Core Design Contradiction:
Quantity of substanceVSSpeed

Solution Approach 1:

The memory system is segmented into multiple storage units with different access speeds. A first storage unit (faster) and a second storage unit (slower but larger capacity) are divided to store different types of data. This segmentation allows the system to optimize between speed and capacity by placing frequently accessed data in the faster unit and less frequently accessed data in the slower unit.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The first storage unit is nested within or coupled to the second storage unit, creating a hierarchical memory structure. The faster storage unit acts as a cache or buffer that sits in front of the larger slower storage unit, allowing the system to access frequently needed data quickly while maintaining large capacity through the nested slower storage unit.

Inventive Principle:
Principle #7Nested doll (Nesting)

2Adaptability or versatility

If intra block copy operations are performed, then video decoding functionality is improved, but pipeline performance deteriorates due to memory latency

Engineering Contradiction:
Improvevideo decoding functionalityVSAvoidpipeline performance
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

Data is prefetched into the first (faster) storage unit before it is actually needed for intra block copy operations. By performing preliminary actions to load data into faster storage in advance, the system avoids memory latency during critical pipeline operations, maintaining high productivity while enabling versatile video decoding functionality.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The first storage unit acts as an intermediary between the slower second storage unit and the video decoding pipeline. It mediates data access by holding frequently accessed data in faster storage, reducing the impact of memory latency on pipeline performance while enabling complex video decoding operations including intra block copy.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentUS10200706B2Pipelined video decoder system
Publication Date: 2019.02.05 AVAGO TECHNOLOGIES INTERNATIONAL SALES PTE LTD
  • US10200706B2 patent drawing
  • US10200706B2 patent drawing
  • US10200706B2 patent drawing

AI summary

Efficient decoding of video content that may involve intra block copy operations, such as copying pixel data from one region of a frame to another region of the same frame is described. For example, a method to decode the video content may involve identifying the video frame in which intra block copy operation is to be performed, prior to the intra block copy operation being initiated. A video decoder may prefetch the pixel data from the source region to a local buffer with low memory latency such that the source pixel data to be copied into the destination blocks in the video frame is readily available. Thus, costly, and time consuming memory access may be avoided, and in turn a video decoding pipeline may operate smoothly without any stalling.