A software profiling tool analyses artificial intelligence workloads on embedded devices, helping developers identify performance bottlenecks, optimise resource usage, and speed up edge AI application development.
Ambiq has introduced heliaPROFILER, an open-source software profiling tool designed to simplify the optimisation of edge AI applications on resource-constrained devices. The tool provides developers with detailed insights into AI workload execution, helping them identify performance bottlenecks and improve model efficiency before deployment.
heliaPROFILER collects execution data from AI inference workloads in the form of execution statistics that include performance of operators, execution time, and resources used. Through providing profiling data, the tool facilitates the identification of inefficient operators, assessment of optimisation opportunities, and performance improvements without having to perform manual analysis.
The tool is compatible with Ambiq’s HELIA edge AI software platform and works within the current embedded development flow. The tool integrates with the company’s AI runtime, compiler, and kernel solutions.
As explained by Ambiq, this tool is designed to help shorten the development time through the provision of data related to performance during the optimisation process. The developers will be able to utilise the collected metrics to optimise the AI models, balance workloads, and optimise the power consumption of AI on edge devices powered by batteries.
This tool helps developers to create better edge AI applications because it provides greater visibility into AI execution on embedded hardware.
