Video Reference Frame Retrieval via Hash Function Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Video data processing apparatuses experience inefficiencies due to frequent TLB misses when accessing multiple large video reference frames, particularly in H.264 encoding schemes, leading to stalled memory access and reduced performance.
Innovation Solution
Incorporating translation circuitry that performs a translation process using video reference frame identifier bits as input values for the hash function, which reduces aliasing and TLB misses by ensuring different entries are indexed for each video reference frame, even when spatial coordinate information is repeated, and allowing for efficient memory access through burst operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple video reference frames are buffered for H.264 encoding, then compression ratio is enhanced, but memory access efficiency deteriorates due to frequent TLB misses
Solution Approach 1:
The virtual address space for video reference frames is segmented by incorporating frame identifier bits into the hash function input, creating distinct hash value segments for each reference frame. This segmentation prevents aliasing between different reference frames in the TLB, allowing each frame to be accessed independently without interfering with others, thus maintaining high memory access efficiency while supporting multiple reference frames for enhanced compression
Solution Approach 2:
The hash function parameters are changed to include video reference frame identifier bits as input values. This parameter modification transforms the hash function from one that only uses spatial coordinates to one that incorporates both spatial coordinates and frame identifiers, fundamentally changing how virtual addresses are mapped to TLB entries and eliminating the aliasing problem that causes frequent TLB misses
2Productivity
If translation circuitry incorporates video reference frame identifier bits in hash function input, then TLB miss frequency is reduced, but device complexity increases
Solution Approach 1:
The translation circuitry is designed to perform multiple functions: it extracts both spatial coordinate information and video reference frame identifier bits from the virtual address, feeds both into the hash function, and generates the translated physical address. This multi-functional design consolidates what could be separate processing stages into a unified translation unit, minimizing the increase in device complexity while achieving the goal of reduced TLB misses
Data Source
AI summary
A video data processing apparatus is provided comprising processing circuitry for performing video processing operations requiring access to video reference frames, and a memory management unit configured to translate virtual addresses into physical addresses. Translation circuitry is provided responsive to a memory access request for reference frame pixel data issued by the processing circuitry to perform a translation process on video reference frame information such that the set of input values for at least one hash function in the memory management unit comprises video reference frame identifier bits contained with the video reference frame information. This approach has been found to reduce the frequency of aliasing in the memory management unit when retrieving video reference frames.