Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore a technical conference talk examining GPU core affinity strategies for managing concurrent workloads on ARM Mali GPUs. Learn about the challenges XR compositors face when executing time-sensitive GPU operations alongside other rendering workloads, where traditional GPU-level preemption solutions may still introduce unacceptable delays. Discover an alternative approach that leverages multiple independent GPU execution cores to reserve dedicated resources for critical workloads, ensuring they can execute without interference from concurrent operations. Examine a detailed case study implementation using the ARM Mali G610 GPU in a four-core configuration, driven by the open-source PanVK/Panthor Vulkan driver stack. Understand how core reservation techniques can eliminate the variability and risks associated with preemption latency, particularly for applications requiring precise timing such as XR compositing operations that must complete framebuffer updates before scanout deadlines.
