Monitor Implementation in Inclusive LLC Multicore Processors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multi-core processor systems face performance bottlenecks due to high synchronization overhead in thread execution, particularly when using inclusive shared last level caches, as current implementations require expensive self-snoops that increase latency and bandwidth utilization.
Innovation Solution
The implementation of a monitor architecture that leverages the inclusive nature of the last level cache to reduce the need for self-snoops, optimizing monitor setup and release procedures by eliminating unnecessary self-snoops and utilizing caching bridge control logic to manage cache states and transactions efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional monitor implementation is used in inclusive LLC systems, then monitor synchronization can be achieved, but self-snoop transactions increase latency and bandwidth utilization
Solution Approach 1:
The patent extracts the self-snoop operation from the monitor setup procedure by leveraging the inclusive LLC property. Instead of performing a self-snoop to check cache state, the system uses the inclusive nature of the LLC to infer that if the monitor address is present in the LLC, all cores must have invalid cache lines for that address, eliminating the need for self-snoop transactions.
Solution Approach 2:
The patent introduces the inclusive LLC as an intermediary mechanism that mediates between the monitor setup request and the cache coherence state verification. The LLC's inclusive structure serves as a mediator that provides implicit coherence information, replacing the need for direct self-snoop operations.
2Reliability
If self-snoop transactions are performed for monitor setup, then cache coherence can be verified, but system interconnect bandwidth is increased
Solution Approach 1:
The patent removes the self-snoop transaction from the monitor setup path by using the inclusive LLC property. The verification of cache coherence is extracted from the expensive self-snoop operation and replaced with a simple LLC presence check, which consumes minimal interconnect bandwidth.
Solution Approach 2:
The inclusive LLC serves itself to provide coherence verification information. By maintaining an inclusive structure where all cache lines present in any core's cache are also present in the LLC, the system enables automatic coherence verification without requiring additional self-snoop transactions that would consume interconnect bandwidth.
3Measurement precision
If monitor setup includes self-snoop operations, then synchronization accuracy is maintained, but throughput is reduced
Solution Approach 1:
The patent extracts the accuracy verification function from the self-snoop operation. The inclusive LLC structure provides implicit verification of cache coherence state, allowing the system to maintain synchronization accuracy while eliminating the time-consuming self-snoop step that limits throughput.
Solution Approach 2:
The inclusive LLC structure performs preliminary organization of cache lines before monitor setup is needed. By pre-establishing the inclusive relationship between LLC and core caches, the system prepares coherence verification information in advance, enabling fast monitor setup without sacrificing accuracy.
Data Source
AI summary
A method and apparatus to implement monitor primitives when a processor employs an inclusive shared last level cache. By the employing an inclusive last level cache, the processor is almost always able to complete a monitor transaction without requiring self snooping through the system interconnect.