Mali-G78 AE GPU Workload Management Use Case - Mali-G78 AE GPU Workload Management Use Case - WP557

This use case demonstrates the sophisticated workload management capabilities of the Mali-G78AE GPU, by using a combination of partitioning and virtualization for efficient resource allocation and application isolation.

• Scenario Overview:
  • Three virtual machines (VMs) running independent rendering applications with separate Vulkan/OpenGL ES stacks and Mali graphics drivers.
  • The hypervisor layer manages execution and resource allocation, ensuring VM isolation while allowing multiple VMs to share the same GPU partition.
• GPU Partitioning:
  • GPU hardware is divided into two units: Partition 0 assigned to VM 1 and VM3 (using cores 0 and 1) and partition 1 assigned to VM 2 (using cores 2 and 3).
  • Each partition functions as an independent GPU, allowing multiple VMs to share resources while maintaining isolated workloads.
• Support for Multiple VMs: The Mali-G78AE can support up to a maximum of eight virtual machines, providing a highly flexible platform for complex systems, such as automotive safety and non-safety displays.
• Benefits of Configuration:
  • Partitioning ensures applications within each VM operate independently, preventing interference and resource contention, thereby enhancing stability.
  • Multiple VMs can share the same partition (as in the case of VM1 and VM3), optimizing resource usage. However, stronger isolation can be achieved when each VM accesses a separate partition.
  • This setup can cater to different security levels, such as critical safety and non-safety applications, making it ideal for automotive and industrial use cases.
• Example Application Scenario: One VM can handle a complex application like navigation or 3D visualization, while the other manages a simpler interface, such as an instrument cluster.

This use case showcases the ability of the Mali-G78AE GPUs to combine partitioning and virtualization for effective resource management in complex, multiapplication environments. It demonstrates the GPU architecture's flexibility and efficiency in handling diverse workloads while ensuring isolation and potentially supporting varying safety requirements.