Cache Coherence Directory Segmentation for Multiprocessor Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In multiprocessor systems, the large directory capacity leads to increased power consumption, longer query times, and lower query efficiency due to high precision requirements, while a small directory capacity results in low precision and directory overflow issues.
Innovation Solution
A cache coherence management method that reduces directory capacity by querying data separately using two paths for exclusive and shared states, optimizing the directory structure to compress the vector array both horizontally and vertically, and sharing directory entries to avoid redundancy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If directory capacity is increased to maintain high precision, then measurement precision is improved, but use of energy increases and loss of time increases
Solution Approach 1:
The patent segments the directory into two separate tables: a first table for recording exclusive state data and a second table for recording shared state data. This segmentation allows the system to query only the relevant table based on the data state, reducing the effective directory capacity needed while maintaining query precision. The segmentation principle directly resolves the contradiction by dividing the large directory into smaller, specialized sub-directories.
Solution Approach 2:
The patent introduces a new dimension by compressing the vector array both horizontally and vertically, and by adding a state dimension (exclusive/shared) to the directory structure. This dimensional transformation allows the system to organize data more efficiently, reducing the overall directory capacity required while maintaining the ability to precisely locate data through the state-based routing mechanism.
2Measurement precision
If directory capacity is increased to maintain high precision, then measurement precision is improved, but loss of time increases due to longer query times
Solution Approach 1:
By segmenting the directory into two separate tables for exclusive and shared states, the patent reduces the search space for each query. When data is in exclusive state, only the first table needs to be queried; when in shared state, only the second table needs to be queried. This segmentation directly reduces query time while maintaining precision through state-based routing.
Solution Approach 2:
The patent performs preliminary classification of data into exclusive or shared states and organizes them in separate tables before queries occur. This preliminary organization allows for faster retrieval during actual queries, as the system doesn't need to search through the entire directory but can directly access the relevant table based on the data state.
3Use of energy by moving object
If directory capacity is decreased to reduce power consumption, then use of energy is improved, but measurement precision deteriorates
Solution Approach 1:
The patent segments the large directory into two smaller, specialized tables that collectively maintain the precision needed for data location. Each table is optimized for its specific data state (exclusive or shared), allowing the system to use smaller capacity tables while achieving the same effective precision as a larger unified directory would provide.
4Loss of time
If directory capacity is decreased to reduce query time, then loss of time is improved, but measurement precision deteriorates
Solution Approach 1:
By segmenting the directory into state-specific tables, the patent reduces query time through smaller search spaces while maintaining precision through the state-based routing mechanism. The segmentation allows each table to be compact yet precise for its specific purpose.
Solution Approach 2:
The patent applies local quality by optimizing each table's structure and content for its specific function. The first table is optimized for exclusive state data, and the second table is optimized for shared state data. This local optimization ensures that each table achieves maximum query efficiency and precision for its specific data type, rather than requiring a large uniform directory.
5Reliability
If directory capacity is increased, then reliability is improved, but device complexity increases
Solution Approach 1:
The patent segments the directory into two specialized tables, which actually reduces device complexity compared to a large unified directory. Each table has a specific, simple structure optimized for its data type, making the overall system less complex than managing a single large directory with diverse data requirements.
Data Source
AI summary
A cache coherence management method, a node controller, and a multiprocessor system that includes a first table, a second table, a node controller, and at least two nodes, where the node controller determines, in the first table according to address information of data, a first entry, where the first entry includes a first field and a second field. The first field records an occupation status of the data, the second field indicates a node that occupies the data exclusively when the first field includes an exclusive state, and the node controller determines a second entry in the second table according to the address information of the data and the second field when the first field includes a shared state, where the second entry includes a third field, and the third field indicates nodes that share the data.