Snoop Filter Segmentation for Mobile Power Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional snoop filter designs in system-on-chip (SoC) designs face inefficiencies in power consumption and latency due to the need to look up both state and address tag information for all n-ways in the snoop filter table for every coherent request, especially in cases of cache misses or invalid entries.
Innovation Solution
The proposed snoop filter separates state information and address tag information into distinct arrays, allowing independent access and lookups, with the state information checked first to determine if an address tag lookup is necessary, thereby reducing unnecessary lookups and power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional snoop filter design looks up both state and address tag information for all n-ways, then filtering accuracy is maintained, but power consumption increases and latency is added
Solution Approach 1:
The patent segments the snoop filter table into two separate arrays: a state array storing validity/ownership information and address tag arrays storing address information. This segmentation allows the system to perform selective lookups - only accessing the address tag array when the state array indicates a potential match, thereby reducing unnecessary power consumption while maintaining filtering accuracy.
2Reliability
If conventional snoop filter design looks up both state and address tag information for all n-ways, then filtering completeness is ensured, but access latency increases
Solution Approach 1:
The patent performs a preliminary lookup of the state array before accessing the address tag array. The state array contains validity and ownership information that can quickly determine whether a full address tag lookup is necessary. This preliminary action filters out unnecessary lookups in advance, reducing access latency while ensuring filtering completeness by only skipping lookups when the state information already indicates no match is possible.
3Loss of energy
If snoop filter separates state and address tag information into distinct arrays, then power consumption is reduced, but device complexity increases
Solution Approach 1:
The patent divides the traditional unified snoop filter table into separate state array and address tag arrays. While this segmentation increases structural complexity, it enables selective access patterns that significantly reduce power consumption. The state array is compact and accessed for every request, while the larger address tag array is only accessed when necessary, optimizing the trade-off between complexity and energy efficiency.
4Measurement precision
If snoop filter maintains comprehensive tracking information, then coherency accuracy is maintained, but lookup time increases
Solution Approach 1:
The patent segments tracking information into state information (validity, ownership) stored in a separate state array and address information stored in address tag arrays. The state array provides quick preliminary filtering, and only when state information indicates a potential match does the system perform the more time-consuming address tag lookup. This segmentation maintains coherency accuracy by preserving all necessary tracking information while reducing lookup time through selective access.
Data Source
AI summary
A method includes: receiving a coherent request from a requester; looking up a state array of a snoop filter table corresponding to an index identified by the coherent request; determining state information corresponding to the coherent request; and determining to access one or more address tag arrays of the snoop filter table based on one or more of the state information, the requester, and a type of the coherent request.