Traineeship Integrate Telemetry Profiler In Cpu-Gpu Framework H/F Exail

Company Presentation

Exail is a leading industrial player in high-tech solutions for the civil and defence sectors. Exail Robotics Belgium develops and produces unmanned surface/underwater vehicles that detect, identify, and destroy sea mines or explosives. Thanks to our innovative standoff concept adopted by the Belgian and Dutch navies, we can significantly reduce the risks involved in mine hunting. We also develop our own sonars, which enable to view and detect objects on the waterbed.

Task description

Our Sonar projects contain many components from Command & Control, system state collection to heavy data processing. One of these components is our SAS (Synthetic Aperture Sonar) processing stack. This sonar data processing runs multiple concurrent threads that each have a heterogeneous workloads both on the CPU and on the GPU (CUDA).

This all happens on a low power embedded system with limited resources, thereby making performance and stability key. It's hard to get a clear view on what's going on exactly when, which stage is consuming what amount of memory is there an under or overload of the GPU, etc...

To remedy this we want to leverage an open-source off-the-shelf solution for real time remote telemetry and profiling. From the core this profiler supports heterogeneous profiling and give us deep statistics of how our code runs on the embedded platform.

Your task entails integrate this profiler (starting with our SAS processing stack) to map out the resource usage of our processing as a prototype.

Required skills

· Good command of Dutch or French. Can have a conversation in English. You will work in an international environment.

· Knowledge of C++, experience with object-oriented programming.

· Being able to work in a team.

· Take initiative in resolving issues and proposing solutions.

· Experience with cmake, make, gtest and Linux (Ubuntu) is a plus

· Experience with larger software projects (multiple repository, JIRA) is a plus

· Experience with CUDA the parallel computing API/library of NVIDIA is a plus

In Practice

- Duration: flexible, minimum 2 months total.

- Physical presence on site is required. Mouscron is about a 15-minute drive from Kortrijk. By public transport, our offices are 15 minutes by bike or bus from the train station. Occasionally, we can arrange a visit (and transport) to the Ostend factory where our unmanned vehicles are built.