Coherence Model Translator for Heterogeneous SoC Agents
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Integration of heterogeneous processing units with different and incompatible coherence models in systems-on-chip (SoCs) poses challenges, leading to higher design costs and longer development cycles due to the need for selecting compatible, but less optimal, processing units.
Innovation Solution
A translator is used to adapt between different coherence models, intermediating the exchange of coherency requests and responses between agents and a coherence controller, making the coherence models of agents and controllers transparent to each other, allowing them to be designed according to their respective protocols.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If heterogeneous processing units with different coherence models are integrated, then system performance and functionality are improved, but integration complexity and design difficulty increase
Solution Approach 1:
The patent introduces a translator as an intermediary component between processing units with different coherence models and the coherence controller. This translator converts coherence model-specific requests and responses into a standardized format, enabling heterogeneous processing units to communicate without direct compatibility requirements. The translator acts as a buffer that handles the complexity of model differences, allowing the coherence controller to operate with a unified interface.
2Ease of manufacture
If processing units with incompatible coherence models are directly integrated, then design costs and development cycles increase, but if compatible units are selected, then system optimality decreases
Solution Approach 1:
The translator serves as a mediation layer that reconciles incompatible coherence models, enabling the integration of diverse processing units without incurring additional design costs or extending development cycles. By standardizing the interface between processing units and the coherence controller, the system achieves both cost efficiency and unit selection flexibility simultaneously.
Solution Approach 2:
The translator dynamically adjusts coherence model parameters by mapping different cache state models and transaction protocols to a standardized representation. This parameter transformation allows processing units with varying coherence characteristics to be integrated seamlessly, providing design flexibility without compromising system coherence.
3Device complexity
If a standardized coherence model is enforced across all agents, then integration is simplified, but agent design freedom and optimization are reduced
Solution Approach 1:
The translator enables agents to maintain their native coherence model designs while interfacing with the standardized coherence controller. This architecture preserves agent design freedom by allowing each agent to implement its optimal coherence model, while the translator handles the adaptation to the standardized interface, thus maintaining integration simplicity.
Data Source
AI summary
A system and method are disclosed for multiple coherent caches supporting agents that use different, incompatible coherence models. Compatibility is implemented by translators that accept coherency requests and snoop responses from an agent and accept snoop requests and coherency responses from a coherence controller. The translators issue corresponding coherency requests and snoop responses to the coherence controller and issue corresponding coherency responses and snoop requests to the agent. Interaction between translators and the coherence controller accord with a generic coherence model, which may be a subset, superset, or partially inclusive of features of any native coherence model. A generic coherence protocol may include binary values for each of characteristics: valid or invalid, owned or non-owned, unique or shared, and clean or dirty.