NARX Model Predicts MPSoC Task Contention Delays
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multi-core processors face challenges in predicting and managing contention delays in shared resources, which is critical for safety-critical systems like avionics, due to the uncertainty of execution time and interference between tasks, making it difficult to ensure determinism and certification.
Innovation Solution
A computer-implemented method using a Machine Learning-based Task Contention Model (ML-TCM) that predicts time-series delays by capturing fine-grain snapshots of performance monitoring counters and employing a Non-linear AutoRegressor with exogenous inputs (NARX) to infer and compensate for missing values, providing accurate characterization and optimization of task scheduling and mapping.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple tasks are executed in parallel on shared resources in MPSoC, then throughput and performance are improved, but contention delays and execution time uncertainty increase
Solution Approach 1:
The patent applies preliminary action by capturing performance monitoring counter (PMC) data during isolated task execution before parallel execution. This pre-captured data serves as training input for the NARX model, enabling prediction of contention delays before actual parallel execution occurs, thus preparing the system in advance to compensate for expected delays
Solution Approach 2:
The patent implements feedback through the NARX model architecture, which uses previously predicted contention delays as input for predicting current delays. The model feeds back historical delay information to improve prediction accuracy for current task execution, creating a closed-loop system that continuously refines its predictions based on past performance
2Measurement precision
If fine-grain instrumentation with multiple snapshots is used to capture PMC data, then measurement precision of contention delays is improved, but device complexity and overhead increase
Solution Approach 1:
The patent applies partial action by capturing PMC data at specific strategic points (entry and exit of tasks, and intermediate snapshots) rather than continuously monitoring all operations. This selective sampling provides sufficient precision for contention delay characterization while avoiding the excessive overhead of comprehensive continuous monitoring
Solution Approach 2:
The patent segments the task execution into distinct phases with snapshots taken at boundaries and key intermediate points. By dividing the monitoring into discrete segments rather than continuous observation, the system achieves precise measurement of contention delays while reducing overall instrumentation complexity and overhead
3Measurement precision
If NARX model with time-series prediction is used to predict contention delays, then prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-training the NARX model offline using PMC data captured during isolated task executions. This pre-training phase prepares the model in advance, so that during actual parallel execution, the model can make predictions with high accuracy without requiring complex real-time computations, thus reducing online computational complexity
Data Source
AI summary
There is provided a computer-implemented method of predicting a timeseries of time delays resulting from contention between tasks running in parallel on a multi-processor system using a trained Machine Learning based Task Contention Model, ML based TCM. The method comprises: executing a plurality of actual execution tasks on the multi-processor system in isolation, and for each task and during execution of the respective task, capturing a timeseries comprising a plurality of uWindows by capturing a plurality of snapshots, each snapshot comprising an array of performance monitoring counters, PMCs, since the previous snapshot, and the time since the previous snapshot. The method also includes inferring, from a time-agnostic ML based regressor, a predicted contention delay for the first uWindow of the timeseries when two or more of the plurality of actual execution tasks are executed on parallel on the multi-processor system given the first captured snapshot for each of the tasks to be completed in parallel. The method further includes inferring, from a Non-linear AutoRegressor with exogenous inputs, NARX, predicted contention delays for each subsequent uWindow of the timeseries based on the respective captured snapshot, and the predicted values for the previous time periods fed back into the NARX.