Ray Allocation for Image Processing Cache Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing apparatuses face performance deterioration due to low cache memory hit ratios and load balancing issues when allocating rays to multiple tree searchers, with traditional methods failing to efficiently manage ray distribution and tree search loads.
Innovation Solution
An image processing apparatus that divides generated rays into subareas based on previous frame tree search information, allocating them to tree searchers to optimize cache efficiency and load balancing by determining the division category with the shortest tree search time and reallocating unfinished rays to completed searches.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If rays are allocated to multiple tree searchers using FIFO method, then device complexity is reduced, but cache memory hit ratio becomes low
Solution Approach 1:
The patent applies preliminary action by pre-dividing the image frame into multiple subareas before ray allocation. This pre-division creates spatial locality groups that will benefit from cache memory, allowing the system to achieve high cache hit ratios without complex dynamic allocation mechanisms during ray processing.
Solution Approach 2:
The patent segments the image frame into multiple subareas and assigns rays within each subarea to the same tree searcher. This segmentation creates groups of rays with spatial locality, ensuring that when one ray accesses cache memory, other rays in the same subarea are likely to hit the cache,从而提高 cache memory hit ratio without requiring complex allocation logic.
2Reliability
If rays are divided into groups for allocation, then cache memory hit ratio is improved, but load balancing deteriorates
Solution Approach 1:
The patent applies dynamics by making the subarea division adaptive rather than fixed. The system dynamically adjusts the number and size of subareas based on the actual distribution of tree search operations, allowing load to be balanced across tree searchers while maintaining the spatial locality needed for high cache hit ratios.
Solution Approach 2:
The patent changes parameters by adjusting the number of subareas and rays per subarea based on workload characteristics. When certain subareas have excessive tree search operations, the system modifies the division to redistribute rays, thereby maintaining both cache efficiency and load balancing across different operating conditions.
3Reliability
If more tree searches are performed in a subarea, then processing completeness is improved, but tree searcher load increases
Solution Approach 1:
The patent segments rays requiring tree searches into different subareas and distributes these subareas across multiple tree searchers. This ensures that no single tree searcher is overwhelmed with all tree search operations, while each subarea maintains enough rays to utilize cache memory effectively, balancing processing completeness with load distribution.
Data Source
AI summary
An image processing apparatus includes a ray generator configured to generate at least one ray, a plurality of tree searchers configured to perform a tree search of the generated at least one ray, an area divider configured to divide the generated at least one ray into a plurality of subareas based on information regarding a number of tree searches of a previous image frame, and a ray allocator configured to allocate the plurality of subareas divided by the area divider to the plurality of tree searchers respectively.