Programmable SoC Performance Characterization for Early Design Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing performance estimation tools for system-on-chips (SoCs) are ineffective in the early stages of system design, as they require fully realized designs and are unable to account for architectural decisions made during this phase.
Innovation Solution
A method using performance characterization techniques, where a characterization database relates parameter settings of the processing system to performance under different traffic profiles, allowing for the generation of a parameter image to set registers and optimize system design without needing a fully realized design.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If performance estimation tools add monitor functionality to existing designs, then performance estimation accuracy is improved, but design stage applicability deteriorates (tools become unusable in early design stages)
Solution Approach 1:
The patent applies preliminary action by pre-characterizing the processing system with monitor functionality integrated into the fabric of the system before actual design development. The performance monitor is built into the processing system architecture itself, allowing performance data collection to begin in early design stages without requiring a fully realized design. This enables performance estimation tools to be used during architectural decision-making phases while maintaining accuracy through the integrated monitor capability.
2Measurement precision
If performance monitor functionality is integrated into the processing system, then performance measurement capability is improved, but system complexity increases
Solution Approach 1:
The patent merges the performance monitor functionality directly into the processing system architecture, combining what would traditionally be separate components into a unified structure. The performance monitor is integrated with the processing elements, interconnect fabric, and memory subsystems, allowing performance data to be collected without requiring separate external monitoring hardware. This merging reduces overall system complexity while maintaining measurement precision.
3Productivity
If comprehensive performance characterization data is collected for all parameter settings, then design optimization capability is improved, but data storage and processing requirements increase
Solution Approach 1:
The patent applies local quality by collecting and storing performance characterization data specific to particular parameter settings and traffic profiles rather than attempting to characterize all possible configurations. The system focuses on capturing performance metrics for relevant operating conditions and parameter combinations, storing this targeted data in a characterization database. This approach provides sufficient design optimization capability while avoiding the exponential growth of data storage requirements that would result from comprehensive full-parameter-space characterization.
Data Source
AI summary
An example method of implementing a system design for a programmable system-on-chip (SOC) having a processing system and programmable logic includes receiving a description of performance objectives for the system design. The method further includes accessing a characterization database that relates parameter settings of the processing system to performance under different traffic profiles as generated by an emulation system comprising the processing system and one or more circuit blocks implemented in the programmable logic. The method further includes obtaining a parameter set from the characterization database based on the description of the performance objectives. The method further includes generating a parameter image for setting registers of the processing system based on the parameter set.