Predictable GPU Wavefront Splitting for Safety-Critical Systems

What is it about?

Wavefront splitting is a technique that reduces the impact of branch divergence, a common performance barrier in graphics processing units (GPUs). We present a new version of this technique that improves the performance of GPU applications while ensuring that tasks can complete within timing constraints in safety-critical applications.

Featured Image

Photo by Remy Gieling on Unsplash

Photo by Remy Gieling on Unsplash

Why is it important?

Designers of safety-critical systems may want to use wavefront splitting to enhance the performance of GPUs. However, they need some guarantees that timing requirements will be met. Our technique provides this performance gain while also reducing the execution time in the worst case scenario.