Vivante Vega IP Enables Full GPU Hardware Virtualization on Mobile and Home Entertainment Devices

SUNNYVALE, Calif., Nov. 6, 2013 /PRNewswire/ — Vivante Corporation today announced another breakthrough in its leading PPA (Power, Performance, and Area) Vega graphics product line. For the first time, full hardware-accelerated GPU virtualization (vGPU) will allow SoC manufacturers to support multiple operating systems (OS) on consumer and embedded platforms for use in mobile, home entertainment, and automotive products.

“Vivante is leading the charge to allow BYOD (work/home) and embedded devices to run split OS scenarios,” said Wei-jin Dai, President and CEO of Vivante. “Now, IT managers can eliminate co-mingling of personal and business data by running both a secure corporate controlled operating environment alongside a potentially less secure OS of the user’s choice, and embedded platforms like automotive infotainment/IVIs can share a single SoC that supports secure and non-secure applications.”

Vega vGPU support has been designed into several silicon products and has undergone extensive validation, testing, and quality assurance (QA) on multiple operating systems including Android 4.4 (KitKat), Linux, Windows Embedded, and several automotive grade RTOSes. Vivante products with built-in vGPU functionality include any standard, off-the-shelf multicore Vega GPUs starting with Vega 2X (GC5000 Series) and higher. Vega cores scale to support multiple client OS instances based on the number of GPU cores available, ranging from two (2) to sixteen (16) independent operating systems instances. The modular core design simplifies place-and-route and silicon layout, making it straightforward to implement any number of Vega cores for the target application.

From an application and software view, Vega vGPU virtualization is transparent and automatically handled by hardware, giving users a responsive interface equivalent to them having multiple dedicated GPUs running their application stack. The vGPU is also targeted to have minimal CPU and system overhead and low latency with features like virtualization assist and hardware address translation for virtual and physical memory addressing. There is also no modification to the hypervisor for the GPU, and no need for the application to process INTs or handle I/O accesses.

To ensure protection of protected resources and the underlying secure OS, the vGPUs have a secure L2 cache and Memory Management Unit (sMMU) which maps directly to the ARM system MMU and supports TrustZone implementations and the CoreLink Cache Coherent Interconnect and Dynamic Memory Controller interface. Each core (and OS) can be sandboxed behind a firewall to create a secure environment where the hypervisor controls all virtualized resources in a protected state.

For more information on Vivante Vega products, please visit http://www.vivantecorp.com/technology/3d.html.