LOCI’s Key Features

System Coverage

Achieve full system coverage using symbol names, eliminating the need for suppliers' source code. This enhances safety and understanding among SOC Cores islands running QM or ASILx software components while also addressing time and performance impacts.

Time and Performance Degradation Prediction

Predict time and performance degradation across different suppliers' versions without needing the source code, allowing for root cause analysis. This approach increases system reliability and quality, ensuring functionality uptime and availability for users. Key features of performance degradation over time: Slowing down functions Function response time Function bottlenecks Function throughput Function energy consumption trends

Test Strategy

Based on behavioral analysis, LOCI Indicate which symbols to focus on during unit and system tests. This increases the coverage of system test scenarios and enhances functional quality.

Power optimization (available Q2, 25)

AI-driven suggestions for reducing energy usage without compromising performance. Track energy consumption across all Functions and tasks, indicating where to optimize energy

Prediction of Performance and Power distribution - even on ELF

Performance

Latency: Time to process inference requests.
Throughput: Number of inference requests processed per second.
Response Time: Time taken to respond to an end-user query.
Execution Time: Time taken to execute an AI model.
Model Inference Time: Time taken for a model to produce output.

Power

Power distribution, Max, Mean,Min per service, and workloads.
Prediction on static data.
Monitoring and trend prediction in real time.
Root cause – where to optimize.

Reliability and Availability (RAS) stack for Inference

Temperature
GPU err
CPU err
Network
Memory
Encryption/Decryption latencies.
Untested code path identification-system level.