ML Prediction of QoR Metrics in Early Circuit Design
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current electronic design automation (EDA) tools lack accurate and reliable methods to predict quality of result (QoR) metrics at early stages of circuit design, leading to time-consuming full-flow runs and inefficient resource allocation, especially for complex circuits like system-on-a-chip (SoC) designs.
Innovation Solution
The implementation of machine learning (ML)-based prediction systems that analyze partial circuit designs to forecast QoR metrics, allowing designers to determine if designs meet constraints early on and rank potential designs for optimal resource allocation, using ML models trained on features from completed design runs to provide fast and accurate predictions for timing, routability, area, power, and memory.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple iterations of full implementation flow are run to observe quality of result (QoR), then accurate QoR feedback is obtained, but the process becomes time-consuming and resource-intensive
Solution Approach 1:
The patent applies preliminary action by training machine learning models in advance using features from early design stages (synthesis, placement, floorplanning) and corresponding QoR metrics from completed implementation flows. The trained models then predict QoR metrics at early stages without requiring full implementation flow execution, enabling designers to assess design quality early and make informed decisions before committing significant computational resources.
Solution Approach 2:
The patent creates a virtual copy of the QoR measurement process through machine learning models that replicate the relationship between early design features and final QoR metrics. Instead of executing the actual full implementation flow to measure QoR, the system uses the trained ML model to generate predicted QoR values that closely approximate what would be obtained from a complete run, significantly reducing measurement time while maintaining accuracy.
2Productivity
If early-stage design exploration is performed with accurate QoR prediction, then resource allocation is optimized, but the prediction system requires significant training data and computational resources
Solution Approach 1:
The patent segments the design process into distinct phases: a development phase where the ML model is trained using features and QoR metrics from completed implementation flows, and an operational phase where the trained model predicts QoR for new designs at early stages. This segmentation allows the system to invest computational resources upfront during training, then achieve high productivity during the operational phase without requiring repeated full implementation flows.
Solution Approach 2:
The patent introduces machine learning models as an intermediary between early design features and final QoR metrics. The ML model learns the complex relationships between design features (such as placement quality, routing density, timing constraints) and QoR outcomes, serving as a mediator that translates early-stage design characteristics into accurate QoR predictions without requiring direct execution of the full implementation flow.
Data Source
AI summary
When designing circuits to meet certain constraint requirements, it is challenging to determine whether a given circuit design will meet the constraints. A designer at an early stage of the circuit design (e.g., synthesis or placement) may have limited information to rely on in order to determine whether the eventual circuit, or some design variation thereof, will satisfy those constraints without fully designing the circuit. The approaches described herein use a machine learning (ML) model to predict, based on features of partial circuit designs at early stages of the design flow, whether the full circuit is likely to meet the constraints. Additionally, the disclosed approaches allow for the ranking of various circuit designs or design implementations to determine best candidates to proceed with the full design.